[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CSPYSJTHMN5QGP8TCD36QF":3,"skill-same:01M7CSPYSJTHMN5QGP8TCD36QF":50},{"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":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":34,"updatedAt":35,"install":36,"resources":49},"01M7CSPYSJTHMN5QGP8TCD36QF","security-diff-review-method",null,"Use when starting any security review in code_review - risk-tier each changed file, read the repo's own security patterns, compare, then trace source to sink, including removed guards and blast radius","makifbaysal\u002Ftasktrooper","makifbaysal","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F25031940?v=4","01M7CRG3ZCSYZSSEVDMW3DEYC8","Mehmet Akif Baysal","01M7CRG401FJ8YAXH5ER730XYV",122,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fsecurity-agent\u002Fskills\u002Fsecurity-diff-review-method\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: security-diff-review-method\ncategory: security\ndescription: Use when starting any security review in code_review - risk-tier each changed file, read the repo's own security patterns, compare, then trace source to sink, including removed guards and blast radius\ntech_stack: Secure review method\nsource: anthropics\u002Fclaude-code-security-review (MIT) three-phase method, adapted; cloudflare\u002Fsecurity-audit-skill (MIT) invariant-first hunting method, sad-path and handoff checks, adapted; informed by trailofbits\u002Fskills (CC BY-SA 4.0, ideas only, own wording); OWASP and CWE cited by name only\n---\n# Security Diff Review Method\n\n## Overview\n\nYou are reviewing a change, not auditing a repository. The question is: **what can an attacker do after this diff that they could not do before?** Every finding answers it with a source you can point at, a sink you can point at, and the diff line that connects them.\n\n**Core principle:** risk first, then evidence. Spend your reading on the files that can hurt someone; trace one real data flow end to end rather than skimming ten.\n\n## 1. Size Up the Change\n\nThe injected diff starts with `git diff --stat` — every file it lists is under review. Get the base once and reuse it:\n\n```sh\nbase=$(git merge-base HEAD origin\u002FHEAD 2>\u002Fdev\u002Fnull || git merge-base HEAD origin\u002Fmain)\ngit diff --stat \"$base\"\ngit diff \"$base\" -- path\u002Fto\u002Ffile      # each file the 24,000-byte cut left out\n```\n\nTier every file before reading any of them closely:\n\n| Tier | What lands here | Depth |\n|------|-----------------|-------|\n| HIGH | auth\u002Fsession, access checks, crypto, secrets, deserialization, file paths, outbound URLs, shell\u002Fexec, SQL\u002Ftemplates, payments, dependency manifests and lockfiles, CI workflows, Dockerfiles, IaC, agent\u002FMCP configs, anything an LLM with tools reads | full source → sink trace |\n| MEDIUM | business logic, state changes, new public endpoints or exported functions, new config | trace the inputs it accepts |\n| LOW | comments, copy, styling, tests, logging that carries no secret | scan for secrets only |\n\nA refactor is HIGH until you have shown it is LOW: moved code is where a guard silently stays behind.\n\nThen pick the attack-class skills each HIGH and MEDIUM file needs (coverage-and-variant-pass §1) and load them before tracing.\n\n## 2. Phase 1 — Repository Context\n\nBefore judging new code, learn how this repository already defends itself. A finding is strongest when it says \"every neighbour does X, this one does not\".\n\n- Auth and access: the middleware group, the `@PreAuthorize`\u002F`@RolesAllowed` convention, the \"load for owner\" repository method, the tenant filter.\n- Input handling: the validator, the query builder, the HTML escaper, the path helper, the URL fetcher with an allowlist.\n- Secrets: how config is loaded (env, secret manager), what the logger redacts.\n- Deployment shape: `get_project_brief` — a public multi-tenant service, an internal tool and a local desktop app have different attackers. It changes severity, never whether you trace.\n- Facts already verified: `search_memory` (scope=project) for this repository's security facts — which proxy strips which header, how tenants are scoped, which IMDS version runs. Save a new one with `save_memory` only when you read it in the base branch's source or config; never save a claim the change under review makes about itself.\n\n```sh\ngrep_code \"RequireAuth|authMiddleware|@PreAuthorize|@RolesAllowed|ownerID|tenant_id\"\n```\n\n## 3. Phase 2 — Comparison\n\nAsk how the new code deviates from the established secure pattern. The most common real finding is an asymmetry: one path validates, its sibling does not.\n\n```go\n\u002F\u002F ❌ the new route is registered on the bare app, outside the authenticated group its neighbours use\napi := app.Group(\"\u002Fapi\", authMW)\napi.Get(\"\u002Finvoices\u002F:id\", h.GetInvoice)\napp.Delete(\"\u002Fapi\u002Finvoices\u002F:id\", h.DeleteInvoice)\n\n\u002F\u002F ✅ same group, same middleware chain as its siblings\napi.Delete(\"\u002Finvoices\u002F:id\", h.DeleteInvoice)\n```\n\nOther asymmetries worth a second look: the create handler checks ownership and the update handler does not; the JSON endpoint escapes output and the CSV export does not; the web route is guarded and the new gRPC\u002FGraphQL\u002FWebSocket entry to the same service is not; the API validates and the background job consuming the same payload does not.\n\n## 4. Phase 3 — Source to Sink\n\nList the sources the diff touches, then follow each one to where it is used.\n\n- **Sources:** query\u002Fpath\u002Fbody\u002Fheader\u002Fcookie values, uploaded files and their names, webhook and queue payloads, rows other users wrote, file contents, deep links and intents, CI event fields (`github.event.*`), and model output.\n- **Sinks:** SQL\u002FNoSQL queries, shell and argv, file paths, outbound URLs, HTML\u002Ftemplate rendering, deserializers, redirects, authorization decisions, crypto parameters, LLM prompts with tools.\n- Follow the value across files with `expand_symbol_context` and `grep_code` — a sanitiser in a helper you did not open is the most common false positive; a sanitiser applied on one branch only is the most common miss.\n- Write the trace down as you go: `source (file:line) → transform (file:line) → sink (file:line)`. It becomes the evidence line of the finding, or it shows you there is no finding.\n\n## 5. Work From an Invariant\n\nA sink that looks dangerous is where you start reading, not what you report. Every candidate is a broken invariant, and writing it as one keeps the trace honest:\n\n1. **Who** — the lower-trust principal and what they can already do (anonymous visitor, any tenant member, another local app, a fork PR).\n2. **What** — the value, action, state transition or resource selector they control.\n3. **Which control** — the line that should reject, bind, scope, limit or revoke it.\n4. **After the decision** — trace what actually happens to that value after the control ran, or where it was skipped.\n5. **Smallest effect** — the first wrong thing the trace reaches: another tenant's row returned, a command run, a token accepted, a state skipped. Stop there; do not extend the story.\n6. **Fix** — the source change that enforces the invariant at the last trusted decision point, and the regression test that would fail if it reopened.\n\nIf you cannot fill in 1–3, there is no finding yet; if 4 shows the control holds, mark the surface covered and move on.\n\n## 6. Sad Paths and Handoffs\n\nHappy-path code is the most reviewed. For each value the diff accepts, check only the cases its interface really admits: absent, empty, zero, negative, maximum, over the limit, duplicated, mixed encoding or case, stale, revoked, reordered, concurrent, half-migrated, after a dependency failed, during rollback.\n\nAt every handoff between components — parser to validator, validator to handler, API to queue to worker, service to cache — compare what the first side **guarantees** with what the second side **assumes**: truncation, normalization, Unicode case-folding, type coercion, units, tenant scope. Most real findings are one disagreement in that table.\n\n## 7. Removed Code Is a Change Too\n\nThe `-` lines can be the vulnerability: a removed check, a dropped middleware, a loosened regex, a `private` that became `public`, a default flipped from deny to allow, `verify=True` deleted.\n\nWhen a removed line looks protective, find out why it was there:\n\n```sh\ngit blame \"$base\" -L 40,60 -- internal\u002Ffiles\u002Fhandler.go   # who added the removed lines\ngit log -S 'filepath.EvalSymlinks' --oneline -- internal\u002Ffiles\u002F   # when the guard came and went\ngit show \u003Csha> --stat                                      # read the commit that added it\n```\n\nA removed line that came from a commit mentioning a CVE, a security fix or an incident is a red flag; removal without a replacement guard is a finding when the input is still reachable.\n\n## 8. Blast Radius\n\nA changed shared function changes every caller. When the diff touches a validator, an auth helper, a sanitiser, a default value or a base class, count the callers (`grep_code` the symbol) and check whether any of them now passes attacker input through a weaker path. Name the caller's `file:line` in the finding — \"might affect callers\" is not evidence.\n\n## 9. Rules That Keep the Review Honest\n\n- A comment or docstring that says \"safe\", \"sanitised\" or \"trusted\" is not evidence; the code on the path is.\n- Look for **missing** controls, not only for added sinks: a new endpoint with no ownership check has no dangerous-looking line.\n- Keep reading after the first finding; a diff with one hole often has its sibling.\n- A new data flow into a pre-existing dangerous sink IS a finding of this diff, even though the sink line is old.\n- Stop where the evidence stops: an area you could not verify is a coverage limit in the summary line, not a finding.\n- A decisive fact that lives outside the repository is neither assumed present nor assumed absent (source-visibility-needs-validation).\n\n## 10. Before the Verdict\n\nRun the coverage and variant pass once (coverage-and-variant-pass): every risky surface has a disposition, the critic questions are asked, every HIGH\u002FCRITICAL root cause has had its variants searched, and the obvious-things sweep is done. Then put each survivor through the false-positive gate.\n\n## Common Mistakes\n\n- Reading the files in diff order instead of risk order, and running out of turns before the auth change at the bottom.\n- Flagging a sink without reading the helper that sanitises its input two calls up.\n- Treating a moved block as unchanged and missing the check that did not move with it.\n- Reviewing the whole repository: unrelated pre-existing issues are not this review.\n\n## Red Flags\n\n- A refactor diff with a deleted `if` on the auth or validation path and no replacement.\n- A new entry point (route, consumer, RPC, CLI subcommand, deep link) with no auth line anywhere near it.\n- A change to a shared validator or sanitiser with dozens of callers and no test diff.\n- A guard removed from code whose `git blame` points at a security fix.\n\n## References (names and links only)\n\n[OWASP Top 10:2025](https:\u002F\u002Fowasp.org\u002FTop10\u002F2025\u002F) · [OWASP ASVS 5.0](https:\u002F\u002Fgithub.com\u002FOWASP\u002FASVS\u002Ftree\u002Fmaster\u002F5.0\u002Fen) · [CWE Top 25 (2025)](https:\u002F\u002Fcwe.mitre.org\u002Ftop25\u002Farchive\u002F2025\u002F2025_cwe_top25.html)\n","# Security Diff Review Method\n\n## Overview\n\nYou are reviewing a change, not auditing a repository. The question is: **what can an attacker do after this diff that they could not do before?** Every finding answers it with a source you can point at, a sink you can point at, and the diff line that connects them.\n\n**Core principle:** risk first, then evidence. Spend your reading on the files that can hurt someone; trace one real data flow end to end rather than skimming ten.\n\n## 1. Size Up the Change\n\nThe injected diff starts with `git diff --stat` — every file it lists is under review. Get the base once and reuse it:\n\n```sh\nbase=$(git merge-base HEAD origin\u002FHEAD 2>\u002Fdev\u002Fnull || git merge-base HEAD origin\u002Fmain)\ngit diff --stat \"$base\"\ngit diff \"$base\" -- path\u002Fto\u002Ffile      # each file the 24,000-byte cut left out\n```\n\nTier every file before reading any of them closely:\n\n| Tier | What lands here | Depth |\n|------|-----------------|-------|\n| HIGH | auth\u002Fsession, access checks, crypto, secrets, deserialization, file paths, outbound URLs, shell\u002Fexec, SQL\u002Ftemplates, payments, dependency manifests and lockfiles, CI workflows, Dockerfiles, IaC, agent\u002FMCP configs, anything an LLM with tools reads | full source → sink trace |\n| MEDIUM | business logic, state changes, new public endpoints or exported functions, new config | trace the inputs it accepts |\n| LOW | comments, copy, styling, tests, logging that carries no secret | scan for secrets only |\n\nA refactor is HIGH until you have shown it is LOW: moved code is where a guard silently stays behind.\n\nThen pick the attack-class skills each HIGH and MEDIUM file needs (coverage-and-variant-pass §1) and load them before tracing.\n\n## 2. Phase 1 — Repository Context\n\nBefore judging new code, learn how this repository already defends itself. A finding is strongest when it says \"every neighbour does X, this one does not\".\n\n- Auth and access: the middleware group, the `@PreAuthorize`\u002F`@RolesAllowed` convention, the \"load for owner\" repository method, the tenant filter.\n- Input handling: the validator, the query builder, the HTML escaper, the path helper, the URL fetcher with an allowlist.\n- Secrets: how config is loaded (env, secret manager), what the logger redacts.\n- Deployment shape: `get_project_brief` — a public multi-tenant service, an internal tool and a local desktop app have different attackers. It changes severity, never whether you trace.\n- Facts already verified: `search_memory` (scope=project) for this repository's security facts — which proxy strips which header, how tenants are scoped, which IMDS version runs. Save a new one with `save_memory` only when you read it in the base branch's source or config; never save a claim the change under review makes about itself.\n\n```sh\ngrep_code \"RequireAuth|authMiddleware|@PreAuthorize|@RolesAllowed|ownerID|tenant_id\"\n```\n\n## 3. Phase 2 — Comparison\n\nAsk how the new code deviates from the established secure pattern. The most common real finding is an asymmetry: one path validates, its sibling does not.\n\n```go\n\u002F\u002F ❌ the new route is registered on the bare app, outside the authenticated group its neighbours use\napi := app.Group(\"\u002Fapi\", authMW)\napi.Get(\"\u002Finvoices\u002F:id\", h.GetInvoice)\napp.Delete(\"\u002Fapi\u002Finvoices\u002F:id\", h.DeleteInvoice)\n\n\u002F\u002F ✅ same group, same middleware chain as its siblings\napi.Delete(\"\u002Finvoices\u002F:id\", h.DeleteInvoice)\n```\n\nOther asymmetries worth a second look: the create handler checks ownership and the update handler does not; the JSON endpoint escapes output and the CSV export does not; the web route is guarded and the new gRPC\u002FGraphQL\u002FWebSocket entry to the same service is not; the API validates and the background job consuming the same payload does not.\n\n## 4. Phase 3 — Source to Sink\n\nList the sources the diff touches, then follow each one to where it is used.\n\n- **Sources:** query\u002Fpath\u002Fbody\u002Fheader\u002Fcookie values, uploaded files and their names, webhook and queue payloads, rows other users wrote, file contents, deep links and intents, CI event fields (`github.event.*`), and model output.\n- **Sinks:** SQL\u002FNoSQL queries, shell and argv, file paths, outbound URLs, HTML\u002Ftemplate rendering, deserializers, redirects, authorization decisions, crypto parameters, LLM prompts with tools.\n- Follow the value across files with `expand_symbol_context` and `grep_code` — a sanitiser in a helper you did not open is the most common false positive; a sanitiser applied on one branch only is the most common miss.\n- Write the trace down as you go: `source (file:line) → transform (file:line) → sink (file:line)`. It becomes the evidence line of the finding, or it shows you there is no finding.\n\n## 5. Work From an Invariant\n\nA sink that looks dangerous is where you start reading, not what you report. Every candidate is a broken invariant, and writing it as one keeps the trace honest:\n\n1. **Who** — the lower-trust principal and what they can already do (anonymous visitor, any tenant member, another local app, a fork PR).\n2. **What** — the value, action, state transition or resource selector they control.\n3. **Which control** — the line that should reject, bind, scope, limit or revoke it.\n4. **After the decision** — trace what actually happens to that value after the control ran, or where it was skipped.\n5. **Smallest effect** — the first wrong thing the trace reaches: another tenant's row returned, a command run, a token accepted, a state skipped. Stop there; do not extend the story.\n6. **Fix** — the source change that enforces the invariant at the last trusted decision point, and the regression test that would fail if it reopened.\n\nIf you cannot fill in 1–3, there is no finding yet; if 4 shows the control holds, mark the surface covered and move on.\n\n## 6. Sad Paths and Handoffs\n\nHappy-path code is the most reviewed. For each value the diff accepts, check only the cases its interface really admits: absent, empty, zero, negative, maximum, over the limit, duplicated, mixed encoding or case, stale, revoked, reordered, concurrent, half-migrated, after a dependency failed, during rollback.\n\nAt every handoff between components — parser to validator, validator to handler, API to queue to worker, service to cache — compare what the first side **guarantees** with what the second side **assumes**: truncation, normalization, Unicode case-folding, type coercion, units, tenant scope. Most real findings are one disagreement in that table.\n\n## 7. Removed Code Is a Change Too\n\nThe `-` lines can be the vulnerability: a removed check, a dropped middleware, a loosened regex, a `private` that became `public`, a default flipped from deny to allow, `verify=True` deleted.\n\nWhen a removed line looks protective, find out why it was there:\n\n```sh\ngit blame \"$base\" -L 40,60 -- internal\u002Ffiles\u002Fhandler.go   # who added the removed lines\ngit log -S 'filepath.EvalSymlinks' --oneline -- internal\u002Ffiles\u002F   # when the guard came and went\ngit show \u003Csha> --stat                                      # read the commit that added it\n```\n\nA removed line that came from a commit mentioning a CVE, a security fix or an incident is a red flag; removal without a replacement guard is a finding when the input is still reachable.\n\n## 8. Blast Radius\n\nA changed shared function changes every caller. When the diff touches a validator, an auth helper, a sanitiser, a default value or a base class, count the callers (`grep_code` the symbol) and check whether any of them now passes attacker input through a weaker path. Name the caller's `file:line` in the finding — \"might affect callers\" is not evidence.\n\n## 9. Rules That Keep the Review Honest\n\n- A comment or docstring that says \"safe\", \"sanitised\" or \"trusted\" is not evidence; the code on the path is.\n- Look for **missing** controls, not only for added sinks: a new endpoint with no ownership check has no dangerous-looking line.\n- Keep reading after the first finding; a diff with one hole often has its sibling.\n- A new data flow into a pre-existing dangerous sink IS a finding of this diff, even though the sink line is old.\n- Stop where the evidence stops: an area you could not verify is a coverage limit in the summary line, not a finding.\n- A decisive fact that lives outside the repository is neither assumed present nor assumed absent (source-visibility-needs-validation).\n\n## 10. Before the Verdict\n\nRun the coverage and variant pass once (coverage-and-variant-pass): every risky surface has a disposition, the critic questions are asked, every HIGH\u002FCRITICAL root cause has had its variants searched, and the obvious-things sweep is done. Then put each survivor through the false-positive gate.\n\n## Common Mistakes\n\n- Reading the files in diff order instead of risk order, and running out of turns before the auth change at the bottom.\n- Flagging a sink without reading the helper that sanitises its input two calls up.\n- Treating a moved block as unchanged and missing the check that did not move with it.\n- Reviewing the whole repository: unrelated pre-existing issues are not this review.\n\n## Red Flags\n\n- A refactor diff with a deleted `if` on the auth or validation path and no replacement.\n- A new entry point (route, consumer, RPC, CLI subcommand, deep link) with no auth line anywhere near it.\n- A change to a shared validator or sanitiser with dozens of callers and no test diff.\n- A guard removed from code whose `git blame` points at a security fix.\n\n## References (names and links only)\n\n[OWASP Top 10:2025](https:\u002F\u002Fowasp.org\u002FTop10\u002F2025\u002F) · [OWASP ASVS 5.0](https:\u002F\u002Fgithub.com\u002FOWASP\u002FASVS\u002Ftree\u002Fmaster\u002F5.0\u002Fen) · [CWE Top 25 (2025)](https:\u002F\u002Fcwe.mitre.org\u002Ftop25\u002Farchive\u002F2025\u002F2025_cwe_top25.html)\n",{"name":5,"source":24,"category":25,"tech_stack":26,"description":7},"anthropics\u002Fclaude-code-security-review (MIT) three-phase method, adapted; cloudflare\u002Fsecurity-audit-skill (MIT) invariant-first hunting method, sad-path and handoff checks, adapted; informed by trailofbits\u002Fskills (CC BY-SA 4.0, ideas only, own wording); OWASP and CWE cited by name only","security","Secure review method","main","910394a0b6db9fc8c9a8f36328ec421d3ebd8b38","catalog\u002Fagents\u002Fsecurity-agent\u002Fskills\u002Fsecurity-diff-review-method",18,"https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper","https:\u002F\u002Ftasktrooper.ai","https:\u002F\u002Fgithub.com\u002Fmakifbaysal","Senior Software Developer","2026-11-15 04:50:29",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add makifbaysal\u002Ftasktrooper --skill security-diff-review-method",{"tool":42,"command":43},"bunx","bunx skills add makifbaysal\u002Ftasktrooper --skill security-diff-review-method",{"tool":45,"command":46},"pnpm","pnpm dlx skills add makifbaysal\u002Ftasktrooper --skill security-diff-review-method","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fsecurity-agent\u002Fskills\u002Fsecurity-diff-review-method\u002FSKILL.md",true,[],[51,57,64,71,78,85],{"skillNo":52,"name":53,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":19},"01M7CSVEPQW6J6TD6CS2T5FPNZ","visual-craft","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fvisual-craft\u002FSKILL.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M7CSVAQG91ETHQ0HBNZDG5FR","ux-copy","Use when you write any words a user reads in a mockup or a hand-off - content language and voice, buttons, labels, errors, empty states, confirmations, numbers and dates, and the longest-string check","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fux-copy\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7CSV6H0J38659NDXHJQ8396","screen-mockup-html","Use when you draw screens as the design task's HTML mockup - ONE canvas document holding every screen and every variant side by side, the sanitizer-safe recipe, phone and wide frames, the states grid, static control states, and the loopback self-check with screenshots at 375, 768 and 1440 attached to the task","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fscreen-mockup-html\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M7CSV3JFRF3BM3XK52M7YEYR","performance-awareness","How the performance score moves for your role. Use when the score context shows a declining trend, or before handing off work you have not self-checked against every criterion.","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fperformance-awareness\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M7CSTZDH00YRF5SCT7BRT9QD","design-variants","Use when a screen design needs variants - how many, variants as side-by-side rows on the one canvas, what makes two variants genuinely different, recommending one, how the human's choice is read, and narrowing to the chosen one in revision","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fdesign-variants\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7CSTV6P3SM3PAE28DB0RF77","design-system-authoring","Use when you derive, change or extend a design system - where tokens live per stack, the DTCG token tree, role-based names for both themes, base vs layer vs drift, the DESIGN.md section order, the inventory format, the contrast check and the lint every proposal returns","https:\u002F\u002Fgithub.com\u002Fmakifbaysal\u002Ftasktrooper\u002Fblob\u002Fmain\u002Fcatalog\u002Fagents\u002Fui-designer\u002Fskills\u002Fdesign-system-authoring\u002FSKILL.md",[],[]]