[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7BJP0X44FE2T3MRCMXVPAKD":3,"skill-same:01M7BJP0X44FE2T3MRCMXVPAKD":47},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7BJP0X44FE2T3MRCMXVPAKD","security-review-methodology",null,"Run a complete, evidence-driven security review of a code repository. Use this Skill whenever the user asks for a security review, vulnerability scan, threat assessment, or penetration-test-style analysis of a codebase — it defines the phase order, the subagent fan-out, the judge triage that removes false positives, and the report format.","ibrahimsaleem\u002Fsaleem-coding-agent","ibrahimsaleem","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F27896839?v=4","01M7AABSBMTHXGZH3XCN8AZNKS","Mohammad Ibrahim Saleem","01M7BJKN6P1XCBK4G79F95VYKP",1,"MIT","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fapps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-security\u002Fskills\u002Fsecurity-review-methodology\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: security-review-methodology\ndescription: Run a complete, evidence-driven security review of a code repository. Use this Skill whenever the user asks for a security review, vulnerability scan, threat assessment, or penetration-test-style analysis of a codebase — it defines the phase order, the subagent fan-out, the judge triage that removes false positives, and the report format.\n---\n\n# Security Review Methodology\n\nFollow these phases in order. Do not skip the scope gate and do not skip the judge phase.\nA finding that never passed a judge does not go in the report.\n\n## Phase 0 — Scope gate\n\nBefore reading a single file, establish and record:\n\n- **What repository**, at what commit or branch.\n- **Authorization**: the user must be entitled to have this code assessed. Ask if it is not obvious.\n- **Runtime scope**: which services you may build and run locally. You may test ONLY instances you\n  launched yourself on this machine. Never point a scan, fuzzer, or exploit at an external host,\n  a third-party API, a staging or production endpoint, or any address the user did not stand up for you.\n- **Depth**: full review, or a named subsystem.\n\nOpen a goal (`goal` tool) for the review so the objective survives compaction. Write the scope into it.\n\nIf authorization or runtime scope is unclear, use `ask_user_question`. Do not guess.\n\n## Phase 1 — Reconnaissance\n\nBuild a map before you look for bugs.\n\n1. Read the manifests (`package.json`, `requirements.txt`, `go.mod`, `pom.xml`, `Cargo.toml`).\n   Note the framework, the auth library, the database driver, the templating engine.\n2. Find the entry points: HTTP route definitions, CLI commands, queue consumers, scheduled jobs,\n   webhook receivers.\n3. Find the trust boundaries: where does untrusted input enter? Where does the code cross into a\n   shell, a SQL statement, a filesystem path, a deserializer, an HTTP client, or a template?\n4. Locate the secrets handling, the session\u002Ftoken logic, and the authorization checks.\n\nRecord the map in your todo list as the scan plan. Recon is read-only — no builds, no execution.\n\n## Phase 2 — Static scan (subagent fan-out)\n\nSplit the codebase into coherent modules (by directory, by route group, or by layer) and dispatch\n**one `subagent` per module**, in parallel. Give each subagent:\n\n- the exact file list for its module,\n- the recon map's trust boundaries,\n- the instruction to report findings as a structured list, each with `file`, `line`, `class`,\n  `severity`, `evidence` (the actual code), and `why-exploitable`.\n\nTell each subagent to look for, at minimum: injection (SQL\u002Fcommand\u002FLDAP\u002Ftemplate), XSS and output\nencoding, authentication and session flaws, authorization gaps and IDOR, SSRF, path traversal and\narbitrary file read\u002Fwrite, insecure deserialization, secrets in source, weak or misused crypto,\nrace conditions in security-relevant paths, and unsafe defaults in configuration.\n\nA subagent that finds nothing reports nothing — that is a valid result, not a failure.\n\n## Phase 3 — Exploit-chain analysis\n\nIndividually-minor findings compose. On the merged finding list, look for chains:\n\n- an information leak that reveals a path, feeding a traversal;\n- a low-privilege write, feeding a code path that executes what it wrote;\n- a CSRF plus a state-changing endpoint;\n- an SSRF plus an internal metadata service.\n\nChains are findings in their own right, and they usually outrank their components in severity.\nRecord each chain with the exact step sequence.\n\n## Phase 4 — Runtime validation\n\nFor every finding whose exploitability is not obvious from the code alone, prove it.\n\n- Build and run the service **locally** — Docker or native, in a background job (`run_in_background`)\n  so the review keeps moving.\n- Write the smallest possible proof: one request, one payload, one observed effect.\n- **Non-destructive only.** Demonstrate impact without destroying data, exhausting resources,\n  installing persistence, or touching anything outside the instance you started.\n- Capture the exact request and the exact response as evidence.\n\nA finding that fails runtime validation is downgraded or dropped — say so explicitly.\n\n## Phase 5 — Judge triage (adversarial)\n\nFor **each** candidate finding, dispatch a `subagent_fork` — it inherits this review's context — with\na single instruction: *argue that this finding is a false positive.* The judge must check:\n\n- Is the sink actually reachable from untrusted input?\n- Is there a sanitizer, a framework-level protection, or a validation layer the scanner missed?\n- Is the \"vulnerable\" code dead, test-only, or behind a flag that is off?\n- Does the claimed severity match the actual impact?\n\nKeep the finding only if the judge fails to kill it. Record the judge's argument and your rebuttal.\nThis phase routinely removes a third of a raw scan's output; that is the point.\n\n## Phase 6 — Report\n\nWrite the report to a file in the workspace. Structure:\n\n1. **Scope** — repository, commit, what was and was not covered, runtime scope.\n2. **Summary** — counts by severity, and the three things that matter most.\n3. **Findings** — one section each, ordered by severity. Every finding carries:\n   title, severity and why that severity, `file:line`, the evidence code, the exploit path or\n   chain, the runtime proof (if any), the judge's challenge and why it failed, and a concrete fix.\n4. **Not vulnerabilities** — findings the judge killed, and why. This section builds trust.\n5. **Coverage gaps** — what you could not assess, and what it would take.\n\nNever include a real credential, token, or piece of personal data you encountered — redact it.\n\n## Phase 7 — Patch branch and pull request (only when asked)\n\nOnly if the user explicitly asks for fixes:\n\n1. Create a branch. Never commit to the default branch.\n2. One commit per finding, each referencing the finding's title.\n3. Re-run the runtime proof against the patched build to show the exploit now fails.\n4. Open the PR with the report's finding sections as the body.\n\n## Working rules\n\n- Use `todo_write` to track phases and per-module scans; a long review outlives your context window.\n- Use `workflow` when a whole-repo sweep is more naturally one orchestrated script than a manual fan-out.\n- Use `ralph` for bounded iterate-until-clean loops (for example, re-scanning after each batch of fixes).\n- Prefer `subagent` for independent module scans and `subagent_fork` for judges that need this review's context.\n- Web search is available for CVE and advisory lookups against the dependency manifests; fetch is off.\n","# Security Review Methodology\n\nFollow these phases in order. Do not skip the scope gate and do not skip the judge phase.\nA finding that never passed a judge does not go in the report.\n\n## Phase 0 — Scope gate\n\nBefore reading a single file, establish and record:\n\n- **What repository**, at what commit or branch.\n- **Authorization**: the user must be entitled to have this code assessed. Ask if it is not obvious.\n- **Runtime scope**: which services you may build and run locally. You may test ONLY instances you\n  launched yourself on this machine. Never point a scan, fuzzer, or exploit at an external host,\n  a third-party API, a staging or production endpoint, or any address the user did not stand up for you.\n- **Depth**: full review, or a named subsystem.\n\nOpen a goal (`goal` tool) for the review so the objective survives compaction. Write the scope into it.\n\nIf authorization or runtime scope is unclear, use `ask_user_question`. Do not guess.\n\n## Phase 1 — Reconnaissance\n\nBuild a map before you look for bugs.\n\n1. Read the manifests (`package.json`, `requirements.txt`, `go.mod`, `pom.xml`, `Cargo.toml`).\n   Note the framework, the auth library, the database driver, the templating engine.\n2. Find the entry points: HTTP route definitions, CLI commands, queue consumers, scheduled jobs,\n   webhook receivers.\n3. Find the trust boundaries: where does untrusted input enter? Where does the code cross into a\n   shell, a SQL statement, a filesystem path, a deserializer, an HTTP client, or a template?\n4. Locate the secrets handling, the session\u002Ftoken logic, and the authorization checks.\n\nRecord the map in your todo list as the scan plan. Recon is read-only — no builds, no execution.\n\n## Phase 2 — Static scan (subagent fan-out)\n\nSplit the codebase into coherent modules (by directory, by route group, or by layer) and dispatch\n**one `subagent` per module**, in parallel. Give each subagent:\n\n- the exact file list for its module,\n- the recon map's trust boundaries,\n- the instruction to report findings as a structured list, each with `file`, `line`, `class`,\n  `severity`, `evidence` (the actual code), and `why-exploitable`.\n\nTell each subagent to look for, at minimum: injection (SQL\u002Fcommand\u002FLDAP\u002Ftemplate), XSS and output\nencoding, authentication and session flaws, authorization gaps and IDOR, SSRF, path traversal and\narbitrary file read\u002Fwrite, insecure deserialization, secrets in source, weak or misused crypto,\nrace conditions in security-relevant paths, and unsafe defaults in configuration.\n\nA subagent that finds nothing reports nothing — that is a valid result, not a failure.\n\n## Phase 3 — Exploit-chain analysis\n\nIndividually-minor findings compose. On the merged finding list, look for chains:\n\n- an information leak that reveals a path, feeding a traversal;\n- a low-privilege write, feeding a code path that executes what it wrote;\n- a CSRF plus a state-changing endpoint;\n- an SSRF plus an internal metadata service.\n\nChains are findings in their own right, and they usually outrank their components in severity.\nRecord each chain with the exact step sequence.\n\n## Phase 4 — Runtime validation\n\nFor every finding whose exploitability is not obvious from the code alone, prove it.\n\n- Build and run the service **locally** — Docker or native, in a background job (`run_in_background`)\n  so the review keeps moving.\n- Write the smallest possible proof: one request, one payload, one observed effect.\n- **Non-destructive only.** Demonstrate impact without destroying data, exhausting resources,\n  installing persistence, or touching anything outside the instance you started.\n- Capture the exact request and the exact response as evidence.\n\nA finding that fails runtime validation is downgraded or dropped — say so explicitly.\n\n## Phase 5 — Judge triage (adversarial)\n\nFor **each** candidate finding, dispatch a `subagent_fork` — it inherits this review's context — with\na single instruction: *argue that this finding is a false positive.* The judge must check:\n\n- Is the sink actually reachable from untrusted input?\n- Is there a sanitizer, a framework-level protection, or a validation layer the scanner missed?\n- Is the \"vulnerable\" code dead, test-only, or behind a flag that is off?\n- Does the claimed severity match the actual impact?\n\nKeep the finding only if the judge fails to kill it. Record the judge's argument and your rebuttal.\nThis phase routinely removes a third of a raw scan's output; that is the point.\n\n## Phase 6 — Report\n\nWrite the report to a file in the workspace. Structure:\n\n1. **Scope** — repository, commit, what was and was not covered, runtime scope.\n2. **Summary** — counts by severity, and the three things that matter most.\n3. **Findings** — one section each, ordered by severity. Every finding carries:\n   title, severity and why that severity, `file:line`, the evidence code, the exploit path or\n   chain, the runtime proof (if any), the judge's challenge and why it failed, and a concrete fix.\n4. **Not vulnerabilities** — findings the judge killed, and why. This section builds trust.\n5. **Coverage gaps** — what you could not assess, and what it would take.\n\nNever include a real credential, token, or piece of personal data you encountered — redact it.\n\n## Phase 7 — Patch branch and pull request (only when asked)\n\nOnly if the user explicitly asks for fixes:\n\n1. Create a branch. Never commit to the default branch.\n2. One commit per finding, each referencing the finding's title.\n3. Re-run the runtime proof against the patched build to show the exploit now fails.\n4. Open the PR with the report's finding sections as the body.\n\n## Working rules\n\n- Use `todo_write` to track phases and per-module scans; a long review outlives your context window.\n- Use `workflow` when a whole-repo sweep is more naturally one orchestrated script than a manual fan-out.\n- Use `ralph` for bounded iterate-until-clean loops (for example, re-scanning after each batch of fixes).\n- Prefer `subagent` for independent module scans and `subagent_fork` for judges that need this review's context.\n- Web search is available for CVE and advisory lookups against the dependency manifests; fetch is off.\n",{"name":5,"description":7},"main","a933d10855d84d7df34f5fcdf6bb5c032328f027","apps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-security\u002Fskills\u002Fsecurity-review-methodology",0,"https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem","https:\u002F\u002Fibrahimsaleem-portfolio.web.app\u002F","Hi Everyone, I am Mohammad Ibrahim Saleem from Houston, USA.\r\nAI Security Engineer @AT&T and Master's in Cybersecurity from University of Houston.","2026-09-12 14:43:33",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add ibrahimsaleem\u002Fsaleem-coding-agent --skill security-review-methodology",{"tool":39,"command":40},"bunx","bunx skills add ibrahimsaleem\u002Fsaleem-coding-agent --skill security-review-methodology",{"tool":42,"command":43},"pnpm","pnpm dlx skills add ibrahimsaleem\u002Fsaleem-coding-agent --skill security-review-methodology","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fapps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-security\u002Fskills\u002Fsecurity-review-methodology\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7BJPC1VJ1MP161V25EXTD5M","user-only-skill","Prove model-disabled skills stay outside the model catalog.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fexamples\u002Facp-agent\u002Ftests\u002Fsnapshots\u002Fskill-load\u002Fworkspace\u002F.dsh\u002Fskills\u002Fuser-only-skill\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"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":19},"01M7BJPA6EV4R4C1S5B5MWBC35","snapshot-skill","Exercise project skill discovery and loading in snapshot tests.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fexamples\u002Facp-agent\u002Ftests\u002Fsnapshots\u002Fskill-load\u002Fworkspace\u002F.dsh\u002Fskills\u002Fsnapshot-skill\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":19},"01M7BJP7N6655321F1DW1D9SXR","model-only-skill","Prove user-disabled skills remain available to the model.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fexamples\u002Facp-agent\u002Ftests\u002Fsnapshots\u002Fskill-load\u002Fworkspace\u002F.dsh\u002Fskills\u002Fmodel-only-skill\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":19},"01M7BJNT3KV38W28SVDKD5WCVR","code-review-methodology","Run a thorough, evidence-driven code review that finds real bugs. Use this Skill whenever the user asks for a code review, a bug hunt, a regression check, a PR review, or an assessment of test coverage — it defines the phase order, the subagent fan-out, the adversarial verification that removes false positives, and the report format.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fapps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-review\u002Fskills\u002Fcode-review-methodology\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":19},"01M7BJNKGA51KSDGYZTJPB62JR","redteam-methodology","Run a structured, contained adversarial evaluation of a target model or AI system. Use this Skill whenever the user asks for red-teaming, safety testing, jailbreak probing, refusal-consistency measurement, or model-behaviour evaluation — it defines the scope gate, probe-suite design, the subagent fan-out, grading, and the report format.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fapps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-redteam\u002Fskills\u002Fredteam-methodology\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":19},"01M7BJNE6NNYZFHK5S207WYZ4F","benchmarking-methodology","Design and run a reproducible benchmark comparing systems, models, or configurations. Use this Skill whenever the user asks for a benchmark, an eval suite, a performance comparison, an A\u002FB of two models or prompts, or a scored task run — it defines the pre-registration step, the run matrix, workflow-driven execution, scoring, and honest statistical reporting.","https:\u002F\u002Fgithub.com\u002Fibrahimsaleem\u002Fsaleem-coding-agent\u002Fblob\u002Fmain\u002Fapps\u002Fcli\u002Fconfig\u002Fagent-presets\u002Fharness-bench\u002Fskills\u002Fbenchmarking-methodology\u002FSKILL.md",[],[]]