[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75XZNHHTFH511MZS5MMJMNW":3,"skill-same:01M75XZNHHTFH511MZS5MMJMNW":51},{"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":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":28,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M75XZNHHTFH511MZS5MMJMNW","audit",null,"Audit current changes, a path, or the full project for quality, security, performance, or test problems and record durable findings. Independent mode prepares or completes a fresh-reviewer checkpoint handoff. Use for \u002Faudit, independent review, security review, code quality review, dead code, duplication, or standards drift.","aiblueprinthq\u002Fai-blueprint","aiblueprinthq","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F317905365?v=4","01M75XZGKSB8GPVXQ69BA57VA7","01M75XZGME4HNZG9CDB8K3TQ8X",510,"MIT","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Faudit\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: audit\ndescription: Audit current changes, a path, or the full project for quality, security, performance, or test problems and record durable findings. Independent mode prepares or completes a fresh-reviewer checkpoint handoff. Use for \u002Faudit, independent review, security review, code quality review, dead code, duplication, or standards drift.\n---\n\n# audit - review code quality against the project standards\n\n**Context reuse:** Reuse any required file already loaded in project instructions or the current session. Read it again only if absent, changed, or exact current bytes or line references are needed.\n\n**First action:** Before project inspection, preflight, or any other tool call,\npublish `running` to `blueprint\u002F.state\u002Frun.json` when `blueprint\u002F.state\u002F`\nexists. From the project root, run the first helper that exists,\n`node .agents\u002Fskills\u002Fblueprint-doctor\u002Fscripts\u002Frun-state.mjs` or\n`node .claude\u002Fskills\u002Fblueprint-doctor\u002Fscripts\u002Frun-state.mjs`, with `start --command\n\u003Cskill> --summary \"\u003Ctruthful summary>\" --boundary\n\u003Cread-only|reviewed|local-only>`. Use `update` at milestones, with `--status\nblocked --resume \u003Ccommand>` for a blocker, and `finish --status ready|completed`\nat the end; `--help` lists every option and `blueprint\u002Fworkflow.md` has the full\nactivity contract. A missing or failing helper is a warning, not a blocker.\n\nWhere this sits in the workflow:\n\n    \u002Fimplement or \u002Fautopilot  ->  [audit]  ->  fixes or \u002Fcomplete\n    (code exists)                 (review +    (repair quality issues\n                                   ledger)      or close the feature)\n\n`\u002Fcheck` proves behavior against the spec. `\u002Fblueprint-doctor` checks Blueprint setup and\nworkflow health. This skill checks the code itself through either a broad review\nor one focused lens: quality, security, performance, or tests.\n\nIt reviews code without changing it: it never edits source files, installs\ndependencies, commits, merges, pushes, or starts product work. A normal audit's\none write is the findings ledger at `blueprint\u002Fcontext\u002Ffindings.md` (Step 4),\nthe durable record of findings and their status. Independent mode may also\nwrite `blueprint\u002Fcontext\u002Freview.md` using the exact record contract in\n`reference\u002Findependent-review.md`.\n\nThe quality-gate config controls when another workflow invokes this skill\nautomatically. An explicit `\u002Faudit` or `$audit` request always selects the audit\nregardless of whether the applicable gate is `manual`, conditional, or `always`.\nA selected `independentReview` gate invokes independent mode instead of letting\nthe builder satisfy its own review. When both audit and independent review are\nselected, one passing independent review satisfies the audit gate.\nA missing config means built-in defaults. If it exists but is invalid, stop and\npoint to `\u002Fblueprint-doctor` before writing the findings ledger.\n\n## Input\n\nTreat scope and lens as separate controls. Arguments may appear in either order,\nsuch as `\u002Faudit security current` or `\u002Faudit src\u002Fauth tests`.\n\nOptional scope:\n\n- no scope argument: use `current` when an active feature exists, otherwise use\n  `changed` when local changes exist, otherwise use `full`\n- `current`: audit the active `current-feature.md`, every committed feature-branch\n  change from its merge base through `HEAD`, staged and unstaged changes,\n  untracked source files, and nearby code affected by the feature\n- `changed`: audit staged, unstaged, and untracked source files plus nearby code\n- `full`: audit all project-owned source, tests, and configuration while excluding\n  dependencies, generated files, build output, coverage output, caches, vendored\n  code, and minified assets unless the user explicitly includes them\n- path or directory: audit that area and the tests or callers needed to understand it\n\nOptional lens:\n\n- no lens: review all four lenses\n- `quality`: maintainability, duplication, dead code, consistency, complexity,\n  and standards drift\n- `security`: authorization, input trust, injection, data exposure, secret\n  handling, and unsafe configuration\n- `performance`: query, network, rendering, memory, payload, concurrency, and\n  unbounded-work risks\n- `tests`: missing coverage for important logic, weak assertions, skipped or\n  focused tests, poor isolation, brittle mocks, and likely flakiness\n\n`full` is always the full-project scope, not a lens. `\u002Faudit full` therefore runs\nall lenses across the full project. When only a lens is supplied, select scope\nwith the normal no-scope rules. A focused pass may name one or more lenses. If\nthe request names multiple lenses, review their union and report them separately.\n\nIf the requested scope is unclear, pick the smallest useful scope and state it.\nIf the lens is unclear, use all lenses and state that choice.\n\nOptional review mode:\n\n- `independent`: prepare or complete an independent review of `current` across\n  all four lenses. It cannot be combined with `changed`, `full`, a path scope,\n  or a focused lens because a completion receipt must cover the whole active\n  work item.\n\n## Independent mode\n\n`\u002Faudit independent current` is a two-context workflow. The builder prepares a\nrequest. With `review.independentExecution: \"manual\"`, the selected fresh\nreviewer session runs the same command to complete it. With `automatic`, the\ncurrent adapter may start a fresh isolated reviewer subagent after preparing the\nrequest. Blueprint verifies the exact target and later staleness. The adapter,\nmodel, and fresh-context identity remain declared metadata.\nAn explicit invocation uses this execution setting even when the active\nworkflow's `independentReview` gate is `manual`; that gate value disables only\nautomatic selection by the workflow.\n\nRead `reference\u002Findependent-review.md` before either phase. It holds the\nrecord contract and both procedures: Phase A prepares the handoff when\n`blueprint\u002Fcontext\u002Freview.md` has no current `pending` request for `HEAD` and\nthe current spec hash; Phase B performs the review when a current pending\nrequest exists.\n\nThe builder never performs Phase B itself. Never let the builder review its\nown work or skip a selected independent-review gate.\n\n## Step 1 - gather context\n\nResolve this required context:\n\n- `AGENTS.md`\n- `blueprint\u002Fconfig.json`\n- `blueprint\u002Fcontext\u002Fproject-overview.md`, targeted: search it for the scope's\n  feature, paths, and distinctive nouns. Read the matching passages plus the\n  security and data passages they touch, meaning the trust, access, and tenancy\n  facts in Users and Usage model and the affected Data model entries. Do not\n  read the whole overview by default.\n- `blueprint\u002Fcontext\u002Fcoding-standards.md`\n- `blueprint\u002Fcontext\u002Fcurrent-feature.md`\n- `blueprint\u002Fcontext\u002Ffindings.md`, for existing IDs and statuses\n- `blueprint\u002Fcontext\u002Freview.md`, for independent request and receipt state\n- `blueprint\u002Fbuild-plan.md`, when feature order matters\n- git branch and working tree status\n- relevant source files, tests, and configs for the chosen scope\n\nFor `current`, `changed`, and path scopes, begin with the diff or named area and\nfollow only the callers, dependencies, tests, and contracts needed to verify a\nreachable finding. Do not survey unrelated directories. For `full`, preserve the\ndeclared exclusions and inspect by bounded area rather than dumping the project\ninto one response.\n\nFor `current`, resolve the comparison base without network access:\n\n1. Use a base branch declared by the active spec or project instructions.\n2. Otherwise use the locally recorded remote default branch when available.\n3. Otherwise use an existing local `main`, then `master`.\n4. Find the merge base and inspect the committed delta through `HEAD`, then add\n   staged, unstaged, and untracked work.\n5. If no reliable base exists, say so and use the active spec plus local changes.\n   Never claim that committed feature work was fully covered in that case.\n\nDo not fetch or pull to discover the base. For `full`, state the excluded paths\nbefore reviewing so generated or third-party code does not consume the audit.\n\nPrefer `rg` and targeted file reads. Do not dump large files into the response.\n\n## Step 2 - run available signals\n\nUse existing commands only. Do not install tools.\n\nRun or inspect only the signals relevant to the selected lens and scope:\n\n- lint and typecheck commands when declared and relevant\n- test command for the tests lens or when it directly validates a suspected risk\n- build command when the selected lens needs compilation or bundle evidence\n- existing security command for the security lens, when declared and locally runnable\n- existing performance command for the performance lens, when declared and locally runnable\n- targeted lightweight searches for the chosen lens, such as unused exports and\n  copied logic for quality, unsafe trust boundaries for security, repeated or\n  unbounded work for performance, and skipped or weak tests for tests\n\nDo not run broad checks unrelated to a focused lens. If a useful command is\nmissing, report that as a gap. Do not invent a pass or claim that a focused\nreview covered the other lenses.\n\n## Step 3 - review the code\n\nFor all lenses, ground findings in reachable code and project-specific\nexpectations. Apply only the selected lens or lenses:\n\n- **Quality:** duplicated logic, dead or unused code, unreachable paths,\n  oversized modules, abstractions that do not pay for themselves, risky missing\n  abstractions, speculative dependencies, services, configuration surfaces,\n  compatibility layers or security machinery, inconsistent patterns, and drift\n  from the standards or spec. Untuned stack-specific template defaults are not\n  established requirements.\n- **Security:** missing authentication or authorization, client-controlled\n  ownership, injection, unsafe parsing or deserialization, sensitive-data\n  exposure, secret handling, insecure defaults, and trust-boundary mistakes.\n  Inspect existing dependency or scanner output when available, but never imply\n  that local manifest inspection is a current vulnerability scan.\n- **Performance:** N+1 queries, repeated network or database work, unnecessary\n  rendering, blocking work on hot paths, unbounded loops or collections, memory\n  growth, oversized payloads, missing pagination, and unsafe concurrency. Mark\n  hypotheses as unverified when runtime or profiling evidence is missing.\n- **Tests:** important logic without coverage when a test command exists, weak\n  assertions, tests that only mirror implementation, excessive mocking, shared\n  state, time or order dependence, skipped or focused tests, placeholder tests,\n  swallowed failures, and missing browser or integration evidence where behavior\n  crosses a real boundary. Never invent a coverage percentage.\n\nDo not nitpick harmless style differences unless they signal drift from the local\npatterns. Prefer a short list of real findings over a broad list of guesses.\n\nFor a proportionality finding, state in **Suggested fix** what can be deleted,\nwhich existing, standard-library, native-platform, or installed mechanism\nreplaces it, and which current requirement would be lost. Use `None` when no\ncurrent requirement would be lost. If the suggested fix removes or changes\nshipped behavior, require an explicit user decision and never describe it as an\nautomatic repair.\n\nDo not broaden a focused pass because another category might be interesting.\nDo not report or call out non-critical concerns from omitted lenses, even as\nsuggestions for a later audit. If an obvious P0 is directly encountered outside\nthe selected lens, report and record it as an out-of-lens critical risk, but do\nnot continue searching that other lens.\n\nIf a possible secret is found, never quote its value, paste the matching source\nline, or include raw command output containing it. Report only the redacted secret\ncategory, file, line, risk, and remediation. Redact sensitive values from all\naudit evidence before responding.\n\n## Step 4 - update the findings ledger\n\n`blueprint\u002Fcontext\u002Ffindings.md` is the durable record of findings. Chat reports\ndo not survive a context clear; the ledger does. It is the only file this skill\nwrites. If it is missing (an older install), create it with a `# Findings`\nheading first.\n\n**The ledger never scopes the review.** Review the code fresh in Step 3, then\nrecord what the review found. Working from the open findings as a checklist and\nverifying only those is the exact failure this file exists to prevent: a repair\ncan introduce a new defect that no existing entry points at.\n\nOne block per finding. The header line is the machine-readable contract and must\nkeep this exact shape; the prose below it is for humans and may vary:\n\n    ### F-03 [P0] open - Retained auth volumes carry the run label\n\n    **File:** ops\u002Fagent-proof\u002Fcompose.yaml:86\n    **Found:** 2026-07-21 by \u002Faudit (scope: current; lens: security)\n    **Why it matters:** ...\n    **Suggested fix:** ...\n    **Resolution:**\n\nIDs are sequential within the ledger (`F-01`, `F-02`, ...), never reused and\nnever renumbered while their entries live here, even after a finding closes.\nBare IDs are scoped to the live ledger: `\u002Fcomplete` archives resolved entries\nunder a work-item prefix (feature 12's first `F-03` becomes `12\u002FF-03`, and its\nsecond build's becomes `12-build-2\u002FF-03`). The build attempt comes from the verified\nspec, not arbitrary filename text; fix prefixes stay their archive filenames. That prefixed form is the permanent reference. A later\nledger that has emptied and reset starts at `F-01` again without colliding. Severity reuses the P0-P3\nscheme from Step 5; only P0 and P1 block `\u002Fcomplete`. Status is one of:\n\n| Status | Meaning | Blocks P0\u002FP1 at \u002Fcomplete |\n|---|---|---|\n| `unverified` | Suspected, no confirming evidence yet | No |\n| `open` | Confirmed, not yet repaired | Yes |\n| `fixed` | Repaired, not yet re-reviewed | Yes |\n| `closed` | Repaired and re-reviewed against the new code | No |\n| `accepted` | Not fixing, by the user's explicit decision; reason recorded in Resolution | No |\n| `invalid` | Re-examination proved the finding wrong; evidence recorded in Resolution | No |\n\nAfter the review:\n\n- Append each new confirmed finding as `open` with the next sequential ID, one\n  past the highest ID present in the ledger (entries carried forward from\n  earlier work count; a fresh ledger starts at `F-01`).\n- Record an unverified risk worth tracking as `unverified`. It is a lead, not a\n  defect, and never gates a merge.\n- Update the entries this pass re-examined: correct the status or severity and\n  note the evidence in **Resolution**.\n- Move a `fixed` finding to `closed` only when all three hold: this pass's\n  reviewed set included the finding's file, re-examining the repaired code\n  confirmed the original defect is gone and the repair introduced no new one,\n  and the report names the finding as closed. An unrelated new finding in the\n  same file gets its own entry and does not keep the repaired one open. Never\n  close a finding implicitly.\n- Set `accepted` only on the user's explicit decision in the current session,\n  and record their reason. Never accept a finding on their behalf.\n- Set `invalid` only when re-examination shows the finding was wrong, and\n  record that evidence in **Resolution**. It is a review verdict (or the\n  user's explicit call), never a shortcut past the gate for blocked work.\n\n`fixed` blocking `\u002Fcomplete` is deliberate: a repair is not done when the code\nchanges, it is done when a review has looked at the result. `\u002Fimplement` marks\nrepairs `fixed`; only a review pass moves them to `closed`.\n\n## Step 5 - report findings\n\nLead with findings, ordered by severity, using the IDs the ledger assigned:\n\n    F-04 [P1] Title\n    File: path:line\n    Why it matters: ...\n    Suggested fix: ...\n\nSeverity:\n\n- `P0` - data loss, security break, or code that cannot ship\n- `P1` - likely bug, broken contract, missing guard, or high-risk duplication\n- `P2` - maintainability issue worth fixing before the feature closes\n- `P3` - small cleanup, consistency issue, or follow-up candidate\n\nUse P0 or P1 only when a concrete code path, violated contract or security\nboundary, failing command or test, or reproducible behavior confirms the risk. If\nthe evidence is incomplete, list it under `Unverified risks` with the missing\nvalidation instead of presenting it as a confirmed high-severity finding.\nAn otherwise pure proportionality finding is P2 or P3. Raise it to P0 or P1 only\nwhen the unnecessary machinery causes a concrete reachable defect or violates an\nestablished security or data-integrity boundary.\n\nIf there are no findings, say that clearly for the selected lens and name any\nremaining risk or missing signal, such as \"no test command declared\" or\n\"browser flow not audited.\"\n\nThen include:\n\n- ledger changes: findings added, updated, or closed this pass, by ID\n- commands run and results\n- selected scope\n- selected lens or lenses\n- base branch, merge base, and commit range for `current`, when available\n- files or directories reviewed\n- generated, third-party, or otherwise excluded paths\n- applicable standards checked\n- browser or runtime evidence inspected, when relevant\n- skipped, focused, or placeholder tests found, when the tests lens was selected\n- checks that were unavailable or could not run\n- suggested repair order\n- independent receipt status and target, when independent mode ran\n\nFor `full`, say whether coverage was complete or partial. Never label a partial\nreview as a full-project audit.\n\n## Rules\n\n- A normal audit writes only the findings ledger. Independent Phase A may also\n  create the exact local spec snapshot and pending request. Phase B may write\n  only `blueprint\u002Fcontext\u002Ffindings.md` and `blueprint\u002Fcontext\u002Freview.md`.\n  Never edit, format, install, commit, merge, push, or delete anything else.\n- Never let a builder complete its own independent request in the same session.\n- Never silently substitute another reviewer adapter or model.\n- A stale receipt is no receipt. Re-review the complete new checkpoint.\n- A focused lens is not a broad audit. State what was not reviewed and never\n  imply that omitted lenses passed.\n- The ledger reports status; it never defines what the review looks at. Do not\n  turn open findings into the review checklist.\n- Never fetch, pull, or run network-backed audit tools without explicit approval.\n- Never reproduce secrets or sensitive values in findings or command output.\n- Findings first. Keep summaries short.\n- Ground every finding in a file path and line number when possible.\n- Avoid speculative rewrites. Recommend the smallest fix that removes the risk.\n- Respect existing project patterns over generic advice.\n- Do not require perfection. The goal is code that is understandable, consistent,\n  testable where it matters, and safe to keep building on.\n\n## Formatting\n\nFormat the output to match the project's conventions in\n`blueprint\u002Fcontext\u002Fai-interaction.md`: concise, scannable markdown, with lists for\nenumerations and tables for matrices rather than dense paragraphs.\n","# audit - review code quality against the project standards\n\n**Context reuse:** Reuse any required file already loaded in project instructions or the current session. Read it again only if absent, changed, or exact current bytes or line references are needed.\n\n**First action:** Before project inspection, preflight, or any other tool call,\npublish `running` to `blueprint\u002F.state\u002Frun.json` when `blueprint\u002F.state\u002F`\nexists. From the project root, run the first helper that exists,\n`node .agents\u002Fskills\u002Fblueprint-doctor\u002Fscripts\u002Frun-state.mjs` or\n`node .claude\u002Fskills\u002Fblueprint-doctor\u002Fscripts\u002Frun-state.mjs`, with `start --command\n\u003Cskill> --summary \"\u003Ctruthful summary>\" --boundary\n\u003Cread-only|reviewed|local-only>`. Use `update` at milestones, with `--status\nblocked --resume \u003Ccommand>` for a blocker, and `finish --status ready|completed`\nat the end; `--help` lists every option and `blueprint\u002Fworkflow.md` has the full\nactivity contract. A missing or failing helper is a warning, not a blocker.\n\nWhere this sits in the workflow:\n\n    \u002Fimplement or \u002Fautopilot  ->  [audit]  ->  fixes or \u002Fcomplete\n    (code exists)                 (review +    (repair quality issues\n                                   ledger)      or close the feature)\n\n`\u002Fcheck` proves behavior against the spec. `\u002Fblueprint-doctor` checks Blueprint setup and\nworkflow health. This skill checks the code itself through either a broad review\nor one focused lens: quality, security, performance, or tests.\n\nIt reviews code without changing it: it never edits source files, installs\ndependencies, commits, merges, pushes, or starts product work. A normal audit's\none write is the findings ledger at `blueprint\u002Fcontext\u002Ffindings.md` (Step 4),\nthe durable record of findings and their status. Independent mode may also\nwrite `blueprint\u002Fcontext\u002Freview.md` using the exact record contract in\n`reference\u002Findependent-review.md`.\n\nThe quality-gate config controls when another workflow invokes this skill\nautomatically. An explicit `\u002Faudit` or `$audit` request always selects the audit\nregardless of whether the applicable gate is `manual`, conditional, or `always`.\nA selected `independentReview` gate invokes independent mode instead of letting\nthe builder satisfy its own review. When both audit and independent review are\nselected, one passing independent review satisfies the audit gate.\nA missing config means built-in defaults. If it exists but is invalid, stop and\npoint to `\u002Fblueprint-doctor` before writing the findings ledger.\n\n## Input\n\nTreat scope and lens as separate controls. Arguments may appear in either order,\nsuch as `\u002Faudit security current` or `\u002Faudit src\u002Fauth tests`.\n\nOptional scope:\n\n- no scope argument: use `current` when an active feature exists, otherwise use\n  `changed` when local changes exist, otherwise use `full`\n- `current`: audit the active `current-feature.md`, every committed feature-branch\n  change from its merge base through `HEAD`, staged and unstaged changes,\n  untracked source files, and nearby code affected by the feature\n- `changed`: audit staged, unstaged, and untracked source files plus nearby code\n- `full`: audit all project-owned source, tests, and configuration while excluding\n  dependencies, generated files, build output, coverage output, caches, vendored\n  code, and minified assets unless the user explicitly includes them\n- path or directory: audit that area and the tests or callers needed to understand it\n\nOptional lens:\n\n- no lens: review all four lenses\n- `quality`: maintainability, duplication, dead code, consistency, complexity,\n  and standards drift\n- `security`: authorization, input trust, injection, data exposure, secret\n  handling, and unsafe configuration\n- `performance`: query, network, rendering, memory, payload, concurrency, and\n  unbounded-work risks\n- `tests`: missing coverage for important logic, weak assertions, skipped or\n  focused tests, poor isolation, brittle mocks, and likely flakiness\n\n`full` is always the full-project scope, not a lens. `\u002Faudit full` therefore runs\nall lenses across the full project. When only a lens is supplied, select scope\nwith the normal no-scope rules. A focused pass may name one or more lenses. If\nthe request names multiple lenses, review their union and report them separately.\n\nIf the requested scope is unclear, pick the smallest useful scope and state it.\nIf the lens is unclear, use all lenses and state that choice.\n\nOptional review mode:\n\n- `independent`: prepare or complete an independent review of `current` across\n  all four lenses. It cannot be combined with `changed`, `full`, a path scope,\n  or a focused lens because a completion receipt must cover the whole active\n  work item.\n\n## Independent mode\n\n`\u002Faudit independent current` is a two-context workflow. The builder prepares a\nrequest. With `review.independentExecution: \"manual\"`, the selected fresh\nreviewer session runs the same command to complete it. With `automatic`, the\ncurrent adapter may start a fresh isolated reviewer subagent after preparing the\nrequest. Blueprint verifies the exact target and later staleness. The adapter,\nmodel, and fresh-context identity remain declared metadata.\nAn explicit invocation uses this execution setting even when the active\nworkflow's `independentReview` gate is `manual`; that gate value disables only\nautomatic selection by the workflow.\n\nRead `reference\u002Findependent-review.md` before either phase. It holds the\nrecord contract and both procedures: Phase A prepares the handoff when\n`blueprint\u002Fcontext\u002Freview.md` has no current `pending` request for `HEAD` and\nthe current spec hash; Phase B performs the review when a current pending\nrequest exists.\n\nThe builder never performs Phase B itself. Never let the builder review its\nown work or skip a selected independent-review gate.\n\n## Step 1 - gather context\n\nResolve this required context:\n\n- `AGENTS.md`\n- `blueprint\u002Fconfig.json`\n- `blueprint\u002Fcontext\u002Fproject-overview.md`, targeted: search it for the scope's\n  feature, paths, and distinctive nouns. Read the matching passages plus the\n  security and data passages they touch, meaning the trust, access, and tenancy\n  facts in Users and Usage model and the affected Data model entries. Do not\n  read the whole overview by default.\n- `blueprint\u002Fcontext\u002Fcoding-standards.md`\n- `blueprint\u002Fcontext\u002Fcurrent-feature.md`\n- `blueprint\u002Fcontext\u002Ffindings.md`, for existing IDs and statuses\n- `blueprint\u002Fcontext\u002Freview.md`, for independent request and receipt state\n- `blueprint\u002Fbuild-plan.md`, when feature order matters\n- git branch and working tree status\n- relevant source files, tests, and configs for the chosen scope\n\nFor `current`, `changed`, and path scopes, begin with the diff or named area and\nfollow only the callers, dependencies, tests, and contracts needed to verify a\nreachable finding. Do not survey unrelated directories. For `full`, preserve the\ndeclared exclusions and inspect by bounded area rather than dumping the project\ninto one response.\n\nFor `current`, resolve the comparison base without network access:\n\n1. Use a base branch declared by the active spec or project instructions.\n2. Otherwise use the locally recorded remote default branch when available.\n3. Otherwise use an existing local `main`, then `master`.\n4. Find the merge base and inspect the committed delta through `HEAD`, then add\n   staged, unstaged, and untracked work.\n5. If no reliable base exists, say so and use the active spec plus local changes.\n   Never claim that committed feature work was fully covered in that case.\n\nDo not fetch or pull to discover the base. For `full`, state the excluded paths\nbefore reviewing so generated or third-party code does not consume the audit.\n\nPrefer `rg` and targeted file reads. Do not dump large files into the response.\n\n## Step 2 - run available signals\n\nUse existing commands only. Do not install tools.\n\nRun or inspect only the signals relevant to the selected lens and scope:\n\n- lint and typecheck commands when declared and relevant\n- test command for the tests lens or when it directly validates a suspected risk\n- build command when the selected lens needs compilation or bundle evidence\n- existing security command for the security lens, when declared and locally runnable\n- existing performance command for the performance lens, when declared and locally runnable\n- targeted lightweight searches for the chosen lens, such as unused exports and\n  copied logic for quality, unsafe trust boundaries for security, repeated or\n  unbounded work for performance, and skipped or weak tests for tests\n\nDo not run broad checks unrelated to a focused lens. If a useful command is\nmissing, report that as a gap. Do not invent a pass or claim that a focused\nreview covered the other lenses.\n\n## Step 3 - review the code\n\nFor all lenses, ground findings in reachable code and project-specific\nexpectations. Apply only the selected lens or lenses:\n\n- **Quality:** duplicated logic, dead or unused code, unreachable paths,\n  oversized modules, abstractions that do not pay for themselves, risky missing\n  abstractions, speculative dependencies, services, configuration surfaces,\n  compatibility layers or security machinery, inconsistent patterns, and drift\n  from the standards or spec. Untuned stack-specific template defaults are not\n  established requirements.\n- **Security:** missing authentication or authorization, client-controlled\n  ownership, injection, unsafe parsing or deserialization, sensitive-data\n  exposure, secret handling, insecure defaults, and trust-boundary mistakes.\n  Inspect existing dependency or scanner output when available, but never imply\n  that local manifest inspection is a current vulnerability scan.\n- **Performance:** N+1 queries, repeated network or database work, unnecessary\n  rendering, blocking work on hot paths, unbounded loops or collections, memory\n  growth, oversized payloads, missing pagination, and unsafe concurrency. Mark\n  hypotheses as unverified when runtime or profiling evidence is missing.\n- **Tests:** important logic without coverage when a test command exists, weak\n  assertions, tests that only mirror implementation, excessive mocking, shared\n  state, time or order dependence, skipped or focused tests, placeholder tests,\n  swallowed failures, and missing browser or integration evidence where behavior\n  crosses a real boundary. Never invent a coverage percentage.\n\nDo not nitpick harmless style differences unless they signal drift from the local\npatterns. Prefer a short list of real findings over a broad list of guesses.\n\nFor a proportionality finding, state in **Suggested fix** what can be deleted,\nwhich existing, standard-library, native-platform, or installed mechanism\nreplaces it, and which current requirement would be lost. Use `None` when no\ncurrent requirement would be lost. If the suggested fix removes or changes\nshipped behavior, require an explicit user decision and never describe it as an\nautomatic repair.\n\nDo not broaden a focused pass because another category might be interesting.\nDo not report or call out non-critical concerns from omitted lenses, even as\nsuggestions for a later audit. If an obvious P0 is directly encountered outside\nthe selected lens, report and record it as an out-of-lens critical risk, but do\nnot continue searching that other lens.\n\nIf a possible secret is found, never quote its value, paste the matching source\nline, or include raw command output containing it. Report only the redacted secret\ncategory, file, line, risk, and remediation. Redact sensitive values from all\naudit evidence before responding.\n\n## Step 4 - update the findings ledger\n\n`blueprint\u002Fcontext\u002Ffindings.md` is the durable record of findings. Chat reports\ndo not survive a context clear; the ledger does. It is the only file this skill\nwrites. If it is missing (an older install), create it with a `# Findings`\nheading first.\n\n**The ledger never scopes the review.** Review the code fresh in Step 3, then\nrecord what the review found. Working from the open findings as a checklist and\nverifying only those is the exact failure this file exists to prevent: a repair\ncan introduce a new defect that no existing entry points at.\n\nOne block per finding. The header line is the machine-readable contract and must\nkeep this exact shape; the prose below it is for humans and may vary:\n\n    ### F-03 [P0] open - Retained auth volumes carry the run label\n\n    **File:** ops\u002Fagent-proof\u002Fcompose.yaml:86\n    **Found:** 2026-07-21 by \u002Faudit (scope: current; lens: security)\n    **Why it matters:** ...\n    **Suggested fix:** ...\n    **Resolution:**\n\nIDs are sequential within the ledger (`F-01`, `F-02`, ...), never reused and\nnever renumbered while their entries live here, even after a finding closes.\nBare IDs are scoped to the live ledger: `\u002Fcomplete` archives resolved entries\nunder a work-item prefix (feature 12's first `F-03` becomes `12\u002FF-03`, and its\nsecond build's becomes `12-build-2\u002FF-03`). The build attempt comes from the verified\nspec, not arbitrary filename text; fix prefixes stay their archive filenames. That prefixed form is the permanent reference. A later\nledger that has emptied and reset starts at `F-01` again without colliding. Severity reuses the P0-P3\nscheme from Step 5; only P0 and P1 block `\u002Fcomplete`. Status is one of:\n\n| Status | Meaning | Blocks P0\u002FP1 at \u002Fcomplete |\n|---|---|---|\n| `unverified` | Suspected, no confirming evidence yet | No |\n| `open` | Confirmed, not yet repaired | Yes |\n| `fixed` | Repaired, not yet re-reviewed | Yes |\n| `closed` | Repaired and re-reviewed against the new code | No |\n| `accepted` | Not fixing, by the user's explicit decision; reason recorded in Resolution | No |\n| `invalid` | Re-examination proved the finding wrong; evidence recorded in Resolution | No |\n\nAfter the review:\n\n- Append each new confirmed finding as `open` with the next sequential ID, one\n  past the highest ID present in the ledger (entries carried forward from\n  earlier work count; a fresh ledger starts at `F-01`).\n- Record an unverified risk worth tracking as `unverified`. It is a lead, not a\n  defect, and never gates a merge.\n- Update the entries this pass re-examined: correct the status or severity and\n  note the evidence in **Resolution**.\n- Move a `fixed` finding to `closed` only when all three hold: this pass's\n  reviewed set included the finding's file, re-examining the repaired code\n  confirmed the original defect is gone and the repair introduced no new one,\n  and the report names the finding as closed. An unrelated new finding in the\n  same file gets its own entry and does not keep the repaired one open. Never\n  close a finding implicitly.\n- Set `accepted` only on the user's explicit decision in the current session,\n  and record their reason. Never accept a finding on their behalf.\n- Set `invalid` only when re-examination shows the finding was wrong, and\n  record that evidence in **Resolution**. It is a review verdict (or the\n  user's explicit call), never a shortcut past the gate for blocked work.\n\n`fixed` blocking `\u002Fcomplete` is deliberate: a repair is not done when the code\nchanges, it is done when a review has looked at the result. `\u002Fimplement` marks\nrepairs `fixed`; only a review pass moves them to `closed`.\n\n## Step 5 - report findings\n\nLead with findings, ordered by severity, using the IDs the ledger assigned:\n\n    F-04 [P1] Title\n    File: path:line\n    Why it matters: ...\n    Suggested fix: ...\n\nSeverity:\n\n- `P0` - data loss, security break, or code that cannot ship\n- `P1` - likely bug, broken contract, missing guard, or high-risk duplication\n- `P2` - maintainability issue worth fixing before the feature closes\n- `P3` - small cleanup, consistency issue, or follow-up candidate\n\nUse P0 or P1 only when a concrete code path, violated contract or security\nboundary, failing command or test, or reproducible behavior confirms the risk. If\nthe evidence is incomplete, list it under `Unverified risks` with the missing\nvalidation instead of presenting it as a confirmed high-severity finding.\nAn otherwise pure proportionality finding is P2 or P3. Raise it to P0 or P1 only\nwhen the unnecessary machinery causes a concrete reachable defect or violates an\nestablished security or data-integrity boundary.\n\nIf there are no findings, say that clearly for the selected lens and name any\nremaining risk or missing signal, such as \"no test command declared\" or\n\"browser flow not audited.\"\n\nThen include:\n\n- ledger changes: findings added, updated, or closed this pass, by ID\n- commands run and results\n- selected scope\n- selected lens or lenses\n- base branch, merge base, and commit range for `current`, when available\n- files or directories reviewed\n- generated, third-party, or otherwise excluded paths\n- applicable standards checked\n- browser or runtime evidence inspected, when relevant\n- skipped, focused, or placeholder tests found, when the tests lens was selected\n- checks that were unavailable or could not run\n- suggested repair order\n- independent receipt status and target, when independent mode ran\n\nFor `full`, say whether coverage was complete or partial. Never label a partial\nreview as a full-project audit.\n\n## Rules\n\n- A normal audit writes only the findings ledger. Independent Phase A may also\n  create the exact local spec snapshot and pending request. Phase B may write\n  only `blueprint\u002Fcontext\u002Ffindings.md` and `blueprint\u002Fcontext\u002Freview.md`.\n  Never edit, format, install, commit, merge, push, or delete anything else.\n- Never let a builder complete its own independent request in the same session.\n- Never silently substitute another reviewer adapter or model.\n- A stale receipt is no receipt. Re-review the complete new checkpoint.\n- A focused lens is not a broad audit. State what was not reviewed and never\n  imply that omitted lenses passed.\n- The ledger reports status; it never defines what the review looks at. Do not\n  turn open findings into the review checklist.\n- Never fetch, pull, or run network-backed audit tools without explicit approval.\n- Never reproduce secrets or sensitive values in findings or command output.\n- Findings first. Keep summaries short.\n- Ground every finding in a file path and line number when possible.\n- Avoid speculative rewrites. Recommend the smallest fix that removes the risk.\n- Respect existing project patterns over generic advice.\n- Do not require perfection. The goal is code that is understandable, consistent,\n  testable where it matters, and safe to keep building on.\n\n## Formatting\n\nFormat the output to match the project's conventions in\n`blueprint\u002Fcontext\u002Fai-interaction.md`: concise, scannable markdown, with lists for\nenumerations and tables for matrices rather than dense paragraphs.\n",{"name":5,"description":7},"main","483eb73dbc58a4b82a5642f864ea7569af947e39",".agents\u002Fskills\u002Faudit",132,"https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint","https:\u002F\u002Fai-blueprint.dev","https:\u002F\u002Fgithub.com\u002Faiblueprinthq","2026-11-14 19:46:05",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add aiblueprinthq\u002Fai-blueprint --skill audit",{"tool":37,"command":38},"bunx","bunx skills add aiblueprinthq\u002Fai-blueprint --skill audit",{"tool":40,"command":41},"pnpm","pnpm dlx skills add aiblueprinthq\u002Fai-blueprint --skill audit","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Faudit\u002FSKILL.md",true,[45],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"reference\u002Findependent-review.md","REFERENCE","参考文档",19455,"https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Faudit\u002Freference\u002Findependent-review.md",[52,58,63,70,77,84],{"skillNo":53,"name":54,"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":55,"categories":56,"keywords":57,"publishedAt":18},"01M75Y34X26C74A66RYE49W8JY","status","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002Fpackages\u002Fcreate-ai-blueprint\u002Ftest\u002Ffixtures\u002Fhealthy-blueprint\u002F.claude\u002Fskills\u002Fstatus\u002FSKILL.md",[],[],{"skillNo":59,"name":54,"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":60,"categories":61,"keywords":62,"publishedAt":18},"01M75Y3456H4WZZ25634QY9KRA","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002Fpackages\u002Fcreate-ai-blueprint\u002Ftest\u002Ffixtures\u002Fhealthy-blueprint\u002F.agents\u002Fskills\u002Fstatus\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":18},"01M75Y31WCYVFM5M4NH7D9MRMW","tests","Set up unit testing with \u002Ftests or \u002Ftests unit, or explicitly set up browser and end-to-end testing with \u002Ftests browser. Reuse or install a suitable test runner, add one example or smoke test, document commands, and verify. Use check for one-time live verification.","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Ftests\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":18},"01M75Y2ZETQZEHN2V19HWZBB1R","prototype","Plan and create throwaway static HTML and CSS mockups with shared design tokens before feature implementation. Use for \u002Fprototype, screen mockups, layout exploration, themes, or deciding a project's look and feel.","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fprototype\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":18},"01M75Y2WY81F9N40G2RADX2FKH","overview","Validate and normalize project-plan.md and build-plan.md, then generate the durable project-overview.md used by agents. Use for \u002Foverview, plan cleanup, generating the first overview, or refreshing context after either plan changes.","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Foverview\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":18},"01M75Y2SNF3FZJP3NHRT0B5MQE","onboard","Onboard a fresh or early scaffold after Blueprint is overlaid by tuning commands, standards, adapters, visibility, and context loading. Use for \u002Fonboard, fresh installation setup, or what to do after installing Blueprint. Use adopt for an established app.","https:\u002F\u002Fgithub.com\u002Faiblueprinthq\u002Fai-blueprint\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fonboard\u002FSKILL.md",[],[]]