[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M736Q0JW7WWNJDE6ZPFAX8HT":3,"skill-same:01M736Q0JW7WWNJDE6ZPFAX8HT":102},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":6,"ownerHtmlUrl":27,"ownerBlogUrl":6,"ownerDescription":28,"updatedAt":29,"install":30,"resources":43},"01M736Q0JW7WWNJDE6ZPFAX8HT","code-review-specialists",null,"Review a code change through an internally selected context and quick, focused, or specialist depth, then return evidence-grounded findings.","graysurf\u002Fagent-runtime-kit","graysurf","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F10785178?v=4","01M736PQHYKHFMNBYF8BT93NM4","01M736PQJWYH7G046WAN7EQCHD",0,"MIT","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: code-review-specialists\ndescription: >\n  Review a code change through an internally selected context and quick,\n  focused, or specialist depth, then return evidence-grounded findings.\n---\n\n# Code Review\n\nUse this as the generic read-only code-review outcome. The workflow selects the\nreview context and the smallest depth that satisfies the request and risk, then\nreturns findings before any summary.\n\n## Contract\n\nPrereqs:\n\n- Run inside the target git repository with `git` available on `PATH`.\n- `review-specialists` is installed from the released nils-cli package and\n  available on `PATH`.\n- Know the base ref for the diff under review, or explicitly choose one before\n  running scope detection.\n- Keep this workflow read-only: it does not auto-fix code, merge, close PRs\u002FMRs,\n  open\u002Fclose issues, or post live provider comments.\n- Dispatch the managed read-only `reviewer-\u003Clens>` subagents for the selected\n  lenses whenever the active host exposes subagent dispatch. In Codex sessions,\n  Codex must dispatch the selected reviewers only when the tool schema exposes\n  a custom-agent selector. The current `multi_agent_v1.spawn_agent` selector,\n  `agent_type`, must be the canonical custom-agent identity from\n  `manifests\u002Fagents.yaml`; `task_name` is only a workflow label and must not\n  stand in for the reviewer identity. If the selector is absent or rejects the\n  requested identity, do not spawn a generic child: report the host capability\n  limitation and use the matching inline fallback. The parent agent owns lens\n  selection, dispatch, validation, and merge.\n- Leave provider review decisions to the owning PR or dispatch parent. Create a\n  `review-evidence` CLI record only when findings need retained evidence.\n\nInputs:\n\n- Diff base ref, optional review target summary, and optional validation\n  evidence to inspect.\n- Optional requested focus or forced specialist flags: `--testing`, `--security`,\n  `--performance`, `--data-migration`, `--api-contract`, `--maintainability`,\n  `--red-team`, or `--all-specialists`.\n- Optional specialist JSONL finding files for deterministic validation, merge,\n  rendering, and bundle synthesis.\n- Optional confidence display threshold for merged findings.\n\nOutputs:\n\n- A context decision (`ad-hoc`, `follow-up`, or `pre-merge`) and a depth\n  decision (`quick`, `focused`, or `specialist`) grounded in request, scope,\n  outer lifecycle, and delivery context.\n- Scope JSON from `review-specialists scope` describing changed files, diff\n  size, stack signals, test framework signals, and suggested specialists.\n- Read-only specialist findings with concrete file or evidence anchors.\n- A final specialist review report using\n  `references\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md`.\n- Optional `review-evidence` records when retained workflow evidence is needed.\n- No source edits, PR\u002FMR comments, merge decisions, or close decisions from this\n  workflow.\n\nFailure modes:\n\n- Base ref is missing or does not resolve in the target repository.\n- Specialist output is malformed JSONL, lacks required fields, uses unsupported\n  severity values, or omits evidence anchors.\n- Findings lack enough confidence or evidence to support a concrete issue; mark\n  them as residual risk instead of presenting them as verified findings.\n- Caller tries to use this read-only workflow as a substitute for provider write\n  authority owned by the PR or dispatch parent,\n  retained evidence, browser-operation checks, CI repair automation, or\n  implementation work.\n\n## Entrypoint\n\nUse the released CLI directly:\n\n```bash\nreview-specialists scope --base \"$BASE_REF\" --format json\nreview-specialists validate --input findings.jsonl --format json\nreview-specialists merge --input findings.jsonl --summary-out specialist-review.md --format json\nreview-specialists render --profile report --input merged-findings.json --out specialist-review.md\nreview-specialists bundle --input findings.jsonl --out-dir \"$REVIEW_OUT\" --profile report --format json\n```\n\n## Mode Selection\n\nThe caller requests the review outcome; the workflow selects context and depth.\n\n- Pre-merge is a delivery context, not a review depth. Every delivered PR still\n  receives a review before merge; the workflow selects quick or full depth\n  inside that context.\n- **Follow-up** — previous findings or review threads are supplied and must be\n  classified as resolved, unresolved, accepted, or residual risk.\n- **Pre-merge** — the review is a delivery gate. Select quick depth only for an\n  eligible L0\u002FL1 routine diff; use the full profile for L2\u002FL3 delivery or any\n  risk trigger. Preserve comment-before-fix ordering and return a decision to\n  the delivery owner.\n- **Focused** — the user explicitly asks for one or more lenses. Force only\n  those lenses unless scope reveals a mandatory safety lens.\n- **Quick** — the diff is small or ordinary, with no migration, security,\n  public API, or other specialist trigger. Dispatch `reviewer-quick` when the\n  active host exposes the custom-agent selector; otherwise inspect the complete\n  diff through the explicit inline fallback. Report concrete findings or a\n  clean result plus residual test risk.\n- **Specialist** — the diff is broad, risky, cross-cutting, or exceeds normal\n  reviewer confidence. Use scope detection and the specialist workflow below.\n- **Quick pre-merge** — quick depth may terminate the delivery review when scope\n  is bounded, validation and checks pass, no suggested or forced risk-specialist\n  trigger or unresolved review state exists, and the outer work is L0 or L1. A `pass` verdict is\n  terminal review evidence for the current head; `findings` blocks merge and\n  enters repair\u002Ffollow-up. A verdict of `escalate` routes to the full pre-merge\n  profile without changing the work tier.\n\nAll modes are read-only. When reviewer subagents are available and a mode uses\none or more lenses, the workflow must dispatch the selected reviewers. The\nparent owns base selection, synthesis, and any authorized provider action.\n\n## Workflow\n\n1. Establish the review target and base ref. For a PR\u002FMR, use the actual\n   PR\u002FMR base or merge-base rather than a moving `origin\u002Fmain` guess. Inspect\n   previous findings and delivery context before selecting a mode.\n2. Select the context and review depth using the precedence above. A caller may\n   prefer quick review, but scope, risk, outer lifecycle, or reviewer confidence\n   can force the full pre-merge profile.\n3. For **quick** mode, dispatch the canonical `reviewer-quick` custom agent\n   when selector-capable dispatch is available. Otherwise state that the\n   selector is unavailable and inspect the complete diff, changed tests, and\n   validation evidence through the inline fallback. Report findings first with\n   file anchors; if clean, state that explicitly and name residual test or\n   validation risk. In quick pre-merge context, return `pass`, `findings`, or\n   `escalate` to the delivery owner; stop without manufacturing specialist work.\n4. For **follow-up** mode, re-check every supplied finding against the current\n   diff and classify it as `resolved`, `unresolved`, `accepted`, or\n   `residual-risk`. Do not broaden scope unless a fix created a new concrete\n   regression.\n5. Run deterministic scope detection for every pre-merge review and for\n   **focused** or **specialist** ad-hoc review:\n\n   ```bash\n   review-specialists scope --base \"$BASE_REF\" --format json\n   ```\n\n6. Select specialists:\n   - In focused mode, use the requested lenses.\n   - In the full pre-merge profile, always include `testing` and\n     `maintainability`.\n   - In specialist mode, use the scope suggestions and rules below.\n   - Always consider `testing` and `maintainability` for larger diffs.\n   - Consider `security` for auth changes or backend changes over 100 diff\n     lines.\n   - Consider `performance` for backend or frontend runtime changes.\n   - Consider `data-migration` for migration, schema, or data transform changes.\n   - Consider `api-contract` for route, controller, API schema, OpenAPI,\n     GraphQL, or protocol changes.\n7. Resolve the matching managed reviewer through the manifest inventory:\n   `reviewer-quick`, `reviewer-testing`, `reviewer-maintainability`,\n   `reviewer-security`, `reviewer-performance`, `reviewer-data-migration`,\n   `reviewer-api-contract`, and `reviewer-red-team`. These are installed at\n   `~\u002F.codex\u002Fagents\u002Freviewer-\u003Clens>.toml` and\n   `~\u002F.claude\u002Fagents\u002Freviewer-\u003Clens>.md`.\n8. Dispatch the selected read-only reviewer subagents one per lens, handing each\n   the base ref and scope; each inspects read-only and returns JSONL findings for\n   its lens. In Codex, set `agent_type` to the canonical custom-agent identity\n   exactly as rendered from the manifest. `task_name` is only a workflow label;\n   an underscore-form task label never selects the profile. If the active tool\n   schema lacks `agent_type`, or the canonical identity is rejected, do not\n   spawn a generic child. State the limitation and run the same lenses through\n   the inline fallback by reading `references\u002Fspecialists\u002F`. You stay the\n   parent: you own base-ref selection, lens selection, dispatch, fallback\n   justification, and the validation\u002Fmerge steps below.\n9. Collect each subagent's JSONL findings (or the inline equivalent) following\n   `references\u002FSPECIALIST_REVIEW_CONTRACT.md`. Treat malformed JSONL, missing\n   required fields, unsupported severities, or absent evidence anchors as a\n   workflow failure or residual risk for that lens — never promote it to a\n   verified finding. Mark unverifiable claims as residual risk, not findings.\n10. Validate and merge findings:\n\n   ```bash\n   review-specialists validate --input findings.jsonl --validate-paths --format json\n   review-specialists merge --input findings.jsonl --summary-out specialist-review.md --format json\n   ```\n\n11. Run red-team only after the selected specialists when `diff_lines > 200`, any\n   selected specialist produced a `critical` finding, or the reviewer forced it.\n   Dispatch `reviewer-red-team` with `agent_type = \"reviewer-red-team\"`\n   when selector-capable subagent dispatch is available. If the selector is\n   unavailable or blocked, state the fallback reason and run the same red-team\n   lens through the inline fallback from `references\u002Fspecialists\u002Fred-team.md`.\n   Hand it the\n   merged first-wave findings so it can probe cross-cutting failure modes. Pass\n   the red-team JSONL through `review-specialists validate`, then append it to\n   the first-wave JSONL and run `review-specialists merge` again over the\n   combined input so duplicate\n   fingerprints, confirming specialists, and confidence ordering are resolved in\n   the final report.\n12. Use the report template for the final synthesis. For either pre-merge\n   profile, apply\n   `references\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md` and\n   `references\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md`; the delivery owner, not a\n   reviewer, posts provider comments and makes the merge decision. For every\n   mode, the recommended next step may route to the dispatch parent's\n   independent-review phase, a normal implementation workflow, or a retained\n   `review-evidence` CLI record, but this workflow does not execute that decision.\n\n## Boundary\n\n`code-review-specialists` owns mode selection, read-only diff inspection, scope\ndetection, specialist selection, reviewer-subagent dispatch, follow-up\nclassification, validation and merge of returned findings, and the final\nreport.\nEach `reviewer-\u003Clens>` subagent owns only its read-only lens.\nThis workflow does not fix code, post PR or MR review comments, mark a draft\nreviewable ready, merge, close issues, or execute the recommended next step —\nthose belong to the owning PR \u002F MR or dispatch parent and the evidence control plane.\n\n## References\n\n- Reviewer subagent sources:\n  `core\u002Fagents\u002Fcode-review\u002Freviewer-\u003Clens>\u002FAGENT.md.tera`\n- Specialist review contract:\n  `references\u002FSPECIALIST_REVIEW_CONTRACT.md`\n- Report template:\n  `references\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md`\n- Specialist prompts:\n  `references\u002Fspecialists\u002F`\n- Delivery specialist review gate:\n  `references\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md`\n- Delivery review outcome comment:\n  `references\u002FDELIVERY_REVIEW_OUTCOME_COMMENT.md`\n- Specialist review comment:\n  `references\u002FSPECIALIST_REVIEW_COMMENT.md`\n- Review outcome posting contract:\n  `references\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md`\n- Delivery review outcome schema:\n  `references\u002FDELIVERY_REVIEW_OUTCOME_SCHEMA.md`\n- Evidence routing policy:\n  `core\u002Fpolicies\u002Fevidence-control-plane.md`\n","# Code Review\n\nUse this as the generic read-only code-review outcome. The workflow selects the\nreview context and the smallest depth that satisfies the request and risk, then\nreturns findings before any summary.\n\n## Contract\n\nPrereqs:\n\n- Run inside the target git repository with `git` available on `PATH`.\n- `review-specialists` is installed from the released nils-cli package and\n  available on `PATH`.\n- Know the base ref for the diff under review, or explicitly choose one before\n  running scope detection.\n- Keep this workflow read-only: it does not auto-fix code, merge, close PRs\u002FMRs,\n  open\u002Fclose issues, or post live provider comments.\n- Dispatch the managed read-only `reviewer-\u003Clens>` subagents for the selected\n  lenses whenever the active host exposes subagent dispatch. In Codex sessions,\n  Codex must dispatch the selected reviewers only when the tool schema exposes\n  a custom-agent selector. The current `multi_agent_v1.spawn_agent` selector,\n  `agent_type`, must be the canonical custom-agent identity from\n  `manifests\u002Fagents.yaml`; `task_name` is only a workflow label and must not\n  stand in for the reviewer identity. If the selector is absent or rejects the\n  requested identity, do not spawn a generic child: report the host capability\n  limitation and use the matching inline fallback. The parent agent owns lens\n  selection, dispatch, validation, and merge.\n- Leave provider review decisions to the owning PR or dispatch parent. Create a\n  `review-evidence` CLI record only when findings need retained evidence.\n\nInputs:\n\n- Diff base ref, optional review target summary, and optional validation\n  evidence to inspect.\n- Optional requested focus or forced specialist flags: `--testing`, `--security`,\n  `--performance`, `--data-migration`, `--api-contract`, `--maintainability`,\n  `--red-team`, or `--all-specialists`.\n- Optional specialist JSONL finding files for deterministic validation, merge,\n  rendering, and bundle synthesis.\n- Optional confidence display threshold for merged findings.\n\nOutputs:\n\n- A context decision (`ad-hoc`, `follow-up`, or `pre-merge`) and a depth\n  decision (`quick`, `focused`, or `specialist`) grounded in request, scope,\n  outer lifecycle, and delivery context.\n- Scope JSON from `review-specialists scope` describing changed files, diff\n  size, stack signals, test framework signals, and suggested specialists.\n- Read-only specialist findings with concrete file or evidence anchors.\n- A final specialist review report using\n  `references\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md`.\n- Optional `review-evidence` records when retained workflow evidence is needed.\n- No source edits, PR\u002FMR comments, merge decisions, or close decisions from this\n  workflow.\n\nFailure modes:\n\n- Base ref is missing or does not resolve in the target repository.\n- Specialist output is malformed JSONL, lacks required fields, uses unsupported\n  severity values, or omits evidence anchors.\n- Findings lack enough confidence or evidence to support a concrete issue; mark\n  them as residual risk instead of presenting them as verified findings.\n- Caller tries to use this read-only workflow as a substitute for provider write\n  authority owned by the PR or dispatch parent,\n  retained evidence, browser-operation checks, CI repair automation, or\n  implementation work.\n\n## Entrypoint\n\nUse the released CLI directly:\n\n```bash\nreview-specialists scope --base \"$BASE_REF\" --format json\nreview-specialists validate --input findings.jsonl --format json\nreview-specialists merge --input findings.jsonl --summary-out specialist-review.md --format json\nreview-specialists render --profile report --input merged-findings.json --out specialist-review.md\nreview-specialists bundle --input findings.jsonl --out-dir \"$REVIEW_OUT\" --profile report --format json\n```\n\n## Mode Selection\n\nThe caller requests the review outcome; the workflow selects context and depth.\n\n- Pre-merge is a delivery context, not a review depth. Every delivered PR still\n  receives a review before merge; the workflow selects quick or full depth\n  inside that context.\n- **Follow-up** — previous findings or review threads are supplied and must be\n  classified as resolved, unresolved, accepted, or residual risk.\n- **Pre-merge** — the review is a delivery gate. Select quick depth only for an\n  eligible L0\u002FL1 routine diff; use the full profile for L2\u002FL3 delivery or any\n  risk trigger. Preserve comment-before-fix ordering and return a decision to\n  the delivery owner.\n- **Focused** — the user explicitly asks for one or more lenses. Force only\n  those lenses unless scope reveals a mandatory safety lens.\n- **Quick** — the diff is small or ordinary, with no migration, security,\n  public API, or other specialist trigger. Dispatch `reviewer-quick` when the\n  active host exposes the custom-agent selector; otherwise inspect the complete\n  diff through the explicit inline fallback. Report concrete findings or a\n  clean result plus residual test risk.\n- **Specialist** — the diff is broad, risky, cross-cutting, or exceeds normal\n  reviewer confidence. Use scope detection and the specialist workflow below.\n- **Quick pre-merge** — quick depth may terminate the delivery review when scope\n  is bounded, validation and checks pass, no suggested or forced risk-specialist\n  trigger or unresolved review state exists, and the outer work is L0 or L1. A `pass` verdict is\n  terminal review evidence for the current head; `findings` blocks merge and\n  enters repair\u002Ffollow-up. A verdict of `escalate` routes to the full pre-merge\n  profile without changing the work tier.\n\nAll modes are read-only. When reviewer subagents are available and a mode uses\none or more lenses, the workflow must dispatch the selected reviewers. The\nparent owns base selection, synthesis, and any authorized provider action.\n\n## Workflow\n\n1. Establish the review target and base ref. For a PR\u002FMR, use the actual\n   PR\u002FMR base or merge-base rather than a moving `origin\u002Fmain` guess. Inspect\n   previous findings and delivery context before selecting a mode.\n2. Select the context and review depth using the precedence above. A caller may\n   prefer quick review, but scope, risk, outer lifecycle, or reviewer confidence\n   can force the full pre-merge profile.\n3. For **quick** mode, dispatch the canonical `reviewer-quick` custom agent\n   when selector-capable dispatch is available. Otherwise state that the\n   selector is unavailable and inspect the complete diff, changed tests, and\n   validation evidence through the inline fallback. Report findings first with\n   file anchors; if clean, state that explicitly and name residual test or\n   validation risk. In quick pre-merge context, return `pass`, `findings`, or\n   `escalate` to the delivery owner; stop without manufacturing specialist work.\n4. For **follow-up** mode, re-check every supplied finding against the current\n   diff and classify it as `resolved`, `unresolved`, `accepted`, or\n   `residual-risk`. Do not broaden scope unless a fix created a new concrete\n   regression.\n5. Run deterministic scope detection for every pre-merge review and for\n   **focused** or **specialist** ad-hoc review:\n\n   ```bash\n   review-specialists scope --base \"$BASE_REF\" --format json\n   ```\n\n6. Select specialists:\n   - In focused mode, use the requested lenses.\n   - In the full pre-merge profile, always include `testing` and\n     `maintainability`.\n   - In specialist mode, use the scope suggestions and rules below.\n   - Always consider `testing` and `maintainability` for larger diffs.\n   - Consider `security` for auth changes or backend changes over 100 diff\n     lines.\n   - Consider `performance` for backend or frontend runtime changes.\n   - Consider `data-migration` for migration, schema, or data transform changes.\n   - Consider `api-contract` for route, controller, API schema, OpenAPI,\n     GraphQL, or protocol changes.\n7. Resolve the matching managed reviewer through the manifest inventory:\n   `reviewer-quick`, `reviewer-testing`, `reviewer-maintainability`,\n   `reviewer-security`, `reviewer-performance`, `reviewer-data-migration`,\n   `reviewer-api-contract`, and `reviewer-red-team`. These are installed at\n   `~\u002F.codex\u002Fagents\u002Freviewer-\u003Clens>.toml` and\n   `~\u002F.claude\u002Fagents\u002Freviewer-\u003Clens>.md`.\n8. Dispatch the selected read-only reviewer subagents one per lens, handing each\n   the base ref and scope; each inspects read-only and returns JSONL findings for\n   its lens. In Codex, set `agent_type` to the canonical custom-agent identity\n   exactly as rendered from the manifest. `task_name` is only a workflow label;\n   an underscore-form task label never selects the profile. If the active tool\n   schema lacks `agent_type`, or the canonical identity is rejected, do not\n   spawn a generic child. State the limitation and run the same lenses through\n   the inline fallback by reading `references\u002Fspecialists\u002F`. You stay the\n   parent: you own base-ref selection, lens selection, dispatch, fallback\n   justification, and the validation\u002Fmerge steps below.\n9. Collect each subagent's JSONL findings (or the inline equivalent) following\n   `references\u002FSPECIALIST_REVIEW_CONTRACT.md`. Treat malformed JSONL, missing\n   required fields, unsupported severities, or absent evidence anchors as a\n   workflow failure or residual risk for that lens — never promote it to a\n   verified finding. Mark unverifiable claims as residual risk, not findings.\n10. Validate and merge findings:\n\n   ```bash\n   review-specialists validate --input findings.jsonl --validate-paths --format json\n   review-specialists merge --input findings.jsonl --summary-out specialist-review.md --format json\n   ```\n\n11. Run red-team only after the selected specialists when `diff_lines > 200`, any\n   selected specialist produced a `critical` finding, or the reviewer forced it.\n   Dispatch `reviewer-red-team` with `agent_type = \"reviewer-red-team\"`\n   when selector-capable subagent dispatch is available. If the selector is\n   unavailable or blocked, state the fallback reason and run the same red-team\n   lens through the inline fallback from `references\u002Fspecialists\u002Fred-team.md`.\n   Hand it the\n   merged first-wave findings so it can probe cross-cutting failure modes. Pass\n   the red-team JSONL through `review-specialists validate`, then append it to\n   the first-wave JSONL and run `review-specialists merge` again over the\n   combined input so duplicate\n   fingerprints, confirming specialists, and confidence ordering are resolved in\n   the final report.\n12. Use the report template for the final synthesis. For either pre-merge\n   profile, apply\n   `references\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md` and\n   `references\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md`; the delivery owner, not a\n   reviewer, posts provider comments and makes the merge decision. For every\n   mode, the recommended next step may route to the dispatch parent's\n   independent-review phase, a normal implementation workflow, or a retained\n   `review-evidence` CLI record, but this workflow does not execute that decision.\n\n## Boundary\n\n`code-review-specialists` owns mode selection, read-only diff inspection, scope\ndetection, specialist selection, reviewer-subagent dispatch, follow-up\nclassification, validation and merge of returned findings, and the final\nreport.\nEach `reviewer-\u003Clens>` subagent owns only its read-only lens.\nThis workflow does not fix code, post PR or MR review comments, mark a draft\nreviewable ready, merge, close issues, or execute the recommended next step —\nthose belong to the owning PR \u002F MR or dispatch parent and the evidence control plane.\n\n## References\n\n- Reviewer subagent sources:\n  `core\u002Fagents\u002Fcode-review\u002Freviewer-\u003Clens>\u002FAGENT.md.tera`\n- Specialist review contract:\n  `references\u002FSPECIALIST_REVIEW_CONTRACT.md`\n- Report template:\n  `references\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md`\n- Specialist prompts:\n  `references\u002Fspecialists\u002F`\n- Delivery specialist review gate:\n  `references\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md`\n- Delivery review outcome comment:\n  `references\u002FDELIVERY_REVIEW_OUTCOME_COMMENT.md`\n- Specialist review comment:\n  `references\u002FSPECIALIST_REVIEW_COMMENT.md`\n- Review outcome posting contract:\n  `references\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md`\n- Delivery review outcome schema:\n  `references\u002FDELIVERY_REVIEW_OUTCOME_SCHEMA.md`\n- Evidence routing policy:\n  `core\u002Fpolicies\u002Fevidence-control-plane.md`\n",{"name":5,"description":7},"main","da7119f2ebece510d6955cea72743069f5496c73","tests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit","https:\u002F\u002Fgithub.com\u002Fgraysurf","Building reliable tools for teams of AI coding agents · Rust \u002F TypeScript · @sympoies","2026-08-29 05:27:47",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add graysurf\u002Fagent-runtime-kit --skill code-review-specialists",{"tool":36,"command":37},"bunx","bunx skills add graysurf\u002Fagent-runtime-kit --skill code-review-specialists",{"tool":39,"command":40},"pnpm","pnpm dlx skills add graysurf\u002Fagent-runtime-kit --skill code-review-specialists","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002FSKILL.md",true,[44,50,54,58,62,66,70,74,78,82,86,90,94,98],{"path":45,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":48,"sourceHtmlUrl":49},"references\u002FDELIVERY_REVIEW_OUTCOME_COMMENT.md","REFERENCE","参考文档",5454,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FDELIVERY_REVIEW_OUTCOME_COMMENT.md",{"path":51,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":52,"sourceHtmlUrl":53},"references\u002FDELIVERY_REVIEW_OUTCOME_SCHEMA.md",1610,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FDELIVERY_REVIEW_OUTCOME_SCHEMA.md",{"path":55,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":56,"sourceHtmlUrl":57},"references\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md",10243,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FDELIVERY_SPECIALIST_REVIEW_GATE.md",{"path":59,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":60,"sourceHtmlUrl":61},"references\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md",17009,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FREVIEW_OUTCOME_POSTING_CONTRACT.md",{"path":63,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":64,"sourceHtmlUrl":65},"references\u002FSPECIALIST_REVIEW_COMMENT.md",3602,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FSPECIALIST_REVIEW_COMMENT.md",{"path":67,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":68,"sourceHtmlUrl":69},"references\u002FSPECIALIST_REVIEW_CONTRACT.md",3456,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FSPECIALIST_REVIEW_CONTRACT.md",{"path":71,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":72,"sourceHtmlUrl":73},"references\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md",1053,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002FSPECIALIST_REVIEW_REPORT_TEMPLATE.md",{"path":75,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":76,"sourceHtmlUrl":77},"references\u002Fspecialists\u002Fapi-contract.md",1314,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fapi-contract.md",{"path":79,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":80,"sourceHtmlUrl":81},"references\u002Fspecialists\u002Fdata-migration.md",1125,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fdata-migration.md",{"path":83,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":84,"sourceHtmlUrl":85},"references\u002Fspecialists\u002Fmaintainability.md",1205,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fmaintainability.md",{"path":87,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":88,"sourceHtmlUrl":89},"references\u002Fspecialists\u002Fperformance.md",1210,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fperformance.md",{"path":91,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":92,"sourceHtmlUrl":93},"references\u002Fspecialists\u002Fred-team.md",1214,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fred-team.md",{"path":95,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":96,"sourceHtmlUrl":97},"references\u002Fspecialists\u002Fsecurity.md",1336,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Fsecurity.md",{"path":99,"resourceType":46,"resourceTypeDesc":47,"sizeBytes":100,"sourceHtmlUrl":101},"references\u002Fspecialists\u002Ftesting.md",1887,"https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Fcode-review\u002Fskills\u002Fcode-review-specialists\u002Fexpected\u002Freferences\u002Fspecialists\u002Ftesting.md",[103,109,114,121,128,135],{"skillNo":104,"name":105,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":18},"01M736XWK14DFRERMQ8RT3HS0Z","removable-skill","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fruntime-smoke\u002Ffixtures\u002Fskill-lifecycle\u002Fremove-skill\u002Ftests\u002Fgolden\u002Fcodex\u002Fplugins\u002Ffixture\u002Fskills\u002Fremovable-skill\u002FSKILL.md",[],[],{"skillNo":110,"name":105,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":18},"01M736XTPQEKAYESDM1REN2KAN","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fruntime-smoke\u002Ffixtures\u002Fskill-lifecycle\u002Fremove-skill\u002Ftests\u002Fgolden\u002Fclaude\u002Fplugins\u002Ffixture\u002Fskills\u002Fremovable-skill\u002FSKILL.md",[],[],{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":118,"categories":119,"keywords":120,"publishedAt":18},"01M736XRMWHMJB4X563G9J6QYY","removable-project-skill","Fixture project-local skill used to prove remove-project-skill dry-run coverage.","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fruntime-smoke\u002Ffixtures\u002Fskill-lifecycle\u002Fremove-project-skill\u002F.agents\u002Fskills\u002Fremovable-project-skill\u002FSKILL.md",[],[],{"skillNo":122,"name":123,"nameZh":6,"description":124,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":127,"publishedAt":18},"01M736XPT65YZMPCBD4K3RHV4P","project-sample-skill","Fixture project-local skill used to prove create-project-skill governance coverage.","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fruntime-smoke\u002Ffixtures\u002Fskill-lifecycle\u002Fcreate-project-skill\u002F.agents\u002Fskills\u002Fproject-sample-skill\u002FSKILL.md",[],[],{"skillNo":129,"name":130,"nameZh":6,"description":131,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":132,"categories":133,"keywords":134,"publishedAt":18},"01M736XKZVYWVQA1YB0P459WND","project-local-skill","Fixture project-local skill used by runtime-kit project-local smoke checks.","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fprojects\u002Fproject-local-smoke\u002F.agents\u002Fskills\u002Fproject-local-skill\u002FSKILL.md",[],[],{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":141,"publishedAt":18},"01M736XHA73KZ5G2TKWSED5P1B","topic-radar","Aggregate read-only AI and technology trend signals (Polymarket, Hacker News, GitHub, arXiv, Hugging Face, news search) into source-grounded Markdown or JSON; the lower-level data engine behind `daily-brief`.","https:\u002F\u002Fgithub.com\u002Fgraysurf\u002Fagent-runtime-kit\u002Fblob\u002Fmain\u002Ftests\u002Fgolden\u002Fhermes\u002Fplugins\u002Freporting\u002Fskills\u002Ftopic-radar\u002Fexpected\u002FSKILL.md",[],[]]