[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CF0KKQZHPK5FBFFRXMX9HT":3,"skill-same:01M7CF0KKQZHPK5FBFFRXMX9HT":49},{"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":6,"updatedAt":34,"install":35,"resources":48},"01M7CF0KKQZHPK5FBFFRXMX9HT","rr-review",null,"One adversarial review of the task's change, at the tier its mode sets — BALANCED in guarded, STRONG in full, security-focused when the domain axis is 2, and in light only when close reports a red fact. A fresh ai-reviewer reads the diff with bounded inputs, returns FINDINGS and a VERDICT, and state.py records them; in-scope BLOCKER and HIGH findings are fixed as one batch, closed by a green verify or one scoped re-review. The pass past the budget is the human's. Use on \"\u002Frr-review\", \"review this\", after \u002Frr-build in guarded or full mode, or when close says a review is required.","vanssata\u002Fclaude-agentic-sdlc-module","vanssata","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1220804?v=4","01M7CEZV36E48TC48Q0V69PFE0","Ivan kakurov","01M7CEZV3SVZE2WWZYMDW13X2X",3,"MIT","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-review\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: rr-review\ndescription: One adversarial review of the task's change, at the tier its mode sets — BALANCED in guarded, STRONG in full, security-focused when the domain axis is 2, and in light only when close reports a red fact. A fresh ai-reviewer reads the diff with bounded inputs, returns FINDINGS and a VERDICT, and state.py records them; in-scope BLOCKER and HIGH findings are fixed as one batch, closed by a green verify or one scoped re-review. The pass past the budget is the human's. Use on \"\u002Frr-review\", \"review this\", after \u002Frr-build in guarded or full mode, or when close says a review is required.\nargument-hint: [B1,H1 … — a scoped re-review of these findings]\n---\n\n# \u002Frr-review $ARGUMENTS\n\nOne reviewer, one pass, on a context that did not write the code. Resolve the\ninstall root once:\n\n```bash\nfor AI_HOME in \"${CLAUDE_CONFIG_DIR:-$HOME\u002F.claude}\" \"${CODEX_HOME:-$HOME\u002F.codex}\"; do\n  [ -d \"$AI_HOME\u002Fskills\u002Frr-review\" ] && break\ndone\nSTATE=\"python3 $AI_HOME\u002Fskills\u002Fai-task\u002Fstate.py\"\n```\n\n## 1. Which review\n\n`$STATE status` gives the mode, the axes and the review so far.\n\n| Mode | Review | Tier |\n|---|---|---|\n| light | only when `close` exits 2 with `review required` (a red fact) | BALANCED |\n| guarded | one | BALANCED |\n| full, full_irreversible | one; briefed with the security focus when the domain axis is 2 | STRONG |\n\n`review` refuses a tier below the mode's. Two passes per task are the budget\n(`review_rounds_max`): one full review and one scoped re-review. The pass past\nit exits 6 `REVIEW_BUDGET` — the next verdict is the human's, from their own\nterminal, with the command it prints, or the pick in the dialog (`DIALOG_PENDING`,\nthen the picked command verbatim, once). Never a third pass. A batch task\n(`\u002Frr-build-all`) has its own budget, `batch.review_rounds_max` full passes\n(shipped three): after a `blockers_open` fix batch the next pass is a full one,\nand the pass past the budget is the human's the same way.\n\n## 2. Brief one `ai-reviewer`\n\nSpawn exactly one `ai-reviewer` at the tier above — under Claude Code with\n`model: $($STATE profile --tier \u003CTIER>)`; under Codex its agent file's model is\nthe tier, so name the tier in the brief. Give it paths, not pasted text:\n\n- the goal, the mode and the workflow (`$STATE status`);\n- the change: `git diff` against the task's base plus new files\n  (`git status --short`), and the task's files from `.ai\u002Freports\u002F\u003Ctask-id>\u002Ftask.md`;\n- the tests the task named, and the last `verify` result and log path;\n- in full mode with domain 2: the security focus — authentication,\n  authorisation, secrets, personal and payment data, input validation,\n  injection — and record the pass with `--focus security`.\n\n**Bounded inputs in light and guarded:** the goal, the diff, the named tests,\nand the project's own threat model (`.ai\u002Fpolicies\u002Fsafety.md`, the guards'\ndefaults, the test fixtures). A finding that needs anything outside them comes\nback `[INFO]` with `in_scope: no`; record it with `$STATE memory risk add`\nas one entry — it never opens a fix. You may move a finding to\n`in_scope: yes`, never the other way. **In full mode** the reviewer's own \"look\nfor\" list applies unbounded. Tell it the cap: about 40 tool calls in light and\nguarded, about 80 in full, then FINDINGS and VERDICT.\n\nBefore you delegate, probe the change yourself for a minute: the few inputs its\nown threat model makes interesting, through the real code, and hand the\nreviewer what you saw — it starts where you stopped. When the change touches a\nguard regex's running time, run the timed probe yourself (`\u002Fusr\u002Fbin\u002Fgrep -E`,\nor Python `timeit` on inputs of at most 4096 chars) and hand over the\nmeasurement: the bounded reviewer may not run timing loops.\n\n## 3. Record the verdict\n\nThe reviewer returns `## FINDINGS` (`- [BLOCKER] B1 …`, `[HIGH] H1`, `[MEDIUM]`,\n`[LOW]`, `[INFO]`, each with `in_scope: yes|no`) and `## VERDICT`\n(`verdict: pass | pass_with_findings | blockers_open`). Pass both through\nverbatim:\n\n```bash\n$STATE review --verdict \u003Cverdict> --tier \u003CBALANCED|STRONG> [--focus security] --report - \u003C\u003C'EOF'\n\u003Cthe reviewer's FINDINGS and VERDICT, unchanged>\nEOF\n```\n\n`review` checks the verdict against the findings: an in-scope BLOCKER is\n`blockers_open`, an in-scope HIGH is not a `pass`. It keeps the text in the\ntask's record (`task.md`), so there is no separate report to write.\n\n## 4. After the verdict\n\nContinue from the verdict §3 recorded; do not record it again. Whatever the\nverdict, each MEDIUM and LOW is one `$STATE memory risk add`\n(`.ai\u002Fproject\u002Fknown-risks.md` amended into the task if missing), not a fix.\n\n- **pass:** `\u002Frr-ship`, i.e. `$STATE close` directly: the verdict is already\n  recorded. (To record and close in one call, §3's `review --verdict pass … \u003C\u003C'EOF' && $STATE close`;\n  `review` exits 0 for every verdict, so that chain is for `pass` only.)\n- **pass_with_findings:** fix every in-scope HIGH in one batch, then\n  `$STATE verify --scope all --after-fix \"H1,H2: \u003Cone line>\"` and\n  `$STATE review --closes \"H1,H2\"`. The findings close on that green run, on a\n  tree the fix changed — no second review. MEDIUM and LOW: `memory risk add`\n  as above; only a BLOCKER, a HIGH or a regression buys an `--after-fix` run.\n- **blockers_open:** fix every BLOCKER and HIGH together, `verify --scope all\n  --after-fix \"…\"`, then **one scoped re-review** (`\u002Frr-review B1,H1`): brief a fresh `ai-reviewer`\n  with the named findings and the fix's diff only — do they close, and what did\n  the fix introduce — and record it as above. Never one finding, one fix, one\n  re-review at a time.\n\nThe `verify` after a fix is often the task's third whole-suite run: past\n`tests.max_suite_runs` it exits 6 `SUITE_BUDGET`, and that run is the human's,\nfrom their own terminal, with the command it prints, or the pick in the dialog.\n\nA finding whose cause is the plan goes back to `\u002Frr-plan` for that part only. A\nreal problem outside the task is recorded with `$STATE memory risk add`; a\nfix in this task closes its risk with `$STATE memory risk close \u003Crkey> --status\nfixed --why \"…\"`. A mistake\nthis repository has seen before — the second time a review flags it — gets one\nline in the project instruction file, in this task.\n","# \u002Frr-review $ARGUMENTS\n\nOne reviewer, one pass, on a context that did not write the code. Resolve the\ninstall root once:\n\n```bash\nfor AI_HOME in \"${CLAUDE_CONFIG_DIR:-$HOME\u002F.claude}\" \"${CODEX_HOME:-$HOME\u002F.codex}\"; do\n  [ -d \"$AI_HOME\u002Fskills\u002Frr-review\" ] && break\ndone\nSTATE=\"python3 $AI_HOME\u002Fskills\u002Fai-task\u002Fstate.py\"\n```\n\n## 1. Which review\n\n`$STATE status` gives the mode, the axes and the review so far.\n\n| Mode | Review | Tier |\n|---|---|---|\n| light | only when `close` exits 2 with `review required` (a red fact) | BALANCED |\n| guarded | one | BALANCED |\n| full, full_irreversible | one; briefed with the security focus when the domain axis is 2 | STRONG |\n\n`review` refuses a tier below the mode's. Two passes per task are the budget\n(`review_rounds_max`): one full review and one scoped re-review. The pass past\nit exits 6 `REVIEW_BUDGET` — the next verdict is the human's, from their own\nterminal, with the command it prints, or the pick in the dialog (`DIALOG_PENDING`,\nthen the picked command verbatim, once). Never a third pass. A batch task\n(`\u002Frr-build-all`) has its own budget, `batch.review_rounds_max` full passes\n(shipped three): after a `blockers_open` fix batch the next pass is a full one,\nand the pass past the budget is the human's the same way.\n\n## 2. Brief one `ai-reviewer`\n\nSpawn exactly one `ai-reviewer` at the tier above — under Claude Code with\n`model: $($STATE profile --tier \u003CTIER>)`; under Codex its agent file's model is\nthe tier, so name the tier in the brief. Give it paths, not pasted text:\n\n- the goal, the mode and the workflow (`$STATE status`);\n- the change: `git diff` against the task's base plus new files\n  (`git status --short`), and the task's files from `.ai\u002Freports\u002F\u003Ctask-id>\u002Ftask.md`;\n- the tests the task named, and the last `verify` result and log path;\n- in full mode with domain 2: the security focus — authentication,\n  authorisation, secrets, personal and payment data, input validation,\n  injection — and record the pass with `--focus security`.\n\n**Bounded inputs in light and guarded:** the goal, the diff, the named tests,\nand the project's own threat model (`.ai\u002Fpolicies\u002Fsafety.md`, the guards'\ndefaults, the test fixtures). A finding that needs anything outside them comes\nback `[INFO]` with `in_scope: no`; record it with `$STATE memory risk add`\nas one entry — it never opens a fix. You may move a finding to\n`in_scope: yes`, never the other way. **In full mode** the reviewer's own \"look\nfor\" list applies unbounded. Tell it the cap: about 40 tool calls in light and\nguarded, about 80 in full, then FINDINGS and VERDICT.\n\nBefore you delegate, probe the change yourself for a minute: the few inputs its\nown threat model makes interesting, through the real code, and hand the\nreviewer what you saw — it starts where you stopped. When the change touches a\nguard regex's running time, run the timed probe yourself (`\u002Fusr\u002Fbin\u002Fgrep -E`,\nor Python `timeit` on inputs of at most 4096 chars) and hand over the\nmeasurement: the bounded reviewer may not run timing loops.\n\n## 3. Record the verdict\n\nThe reviewer returns `## FINDINGS` (`- [BLOCKER] B1 …`, `[HIGH] H1`, `[MEDIUM]`,\n`[LOW]`, `[INFO]`, each with `in_scope: yes|no`) and `## VERDICT`\n(`verdict: pass | pass_with_findings | blockers_open`). Pass both through\nverbatim:\n\n```bash\n$STATE review --verdict \u003Cverdict> --tier \u003CBALANCED|STRONG> [--focus security] --report - \u003C\u003C'EOF'\n\u003Cthe reviewer's FINDINGS and VERDICT, unchanged>\nEOF\n```\n\n`review` checks the verdict against the findings: an in-scope BLOCKER is\n`blockers_open`, an in-scope HIGH is not a `pass`. It keeps the text in the\ntask's record (`task.md`), so there is no separate report to write.\n\n## 4. After the verdict\n\nContinue from the verdict §3 recorded; do not record it again. Whatever the\nverdict, each MEDIUM and LOW is one `$STATE memory risk add`\n(`.ai\u002Fproject\u002Fknown-risks.md` amended into the task if missing), not a fix.\n\n- **pass:** `\u002Frr-ship`, i.e. `$STATE close` directly: the verdict is already\n  recorded. (To record and close in one call, §3's `review --verdict pass … \u003C\u003C'EOF' && $STATE close`;\n  `review` exits 0 for every verdict, so that chain is for `pass` only.)\n- **pass_with_findings:** fix every in-scope HIGH in one batch, then\n  `$STATE verify --scope all --after-fix \"H1,H2: \u003Cone line>\"` and\n  `$STATE review --closes \"H1,H2\"`. The findings close on that green run, on a\n  tree the fix changed — no second review. MEDIUM and LOW: `memory risk add`\n  as above; only a BLOCKER, a HIGH or a regression buys an `--after-fix` run.\n- **blockers_open:** fix every BLOCKER and HIGH together, `verify --scope all\n  --after-fix \"…\"`, then **one scoped re-review** (`\u002Frr-review B1,H1`): brief a fresh `ai-reviewer`\n  with the named findings and the fix's diff only — do they close, and what did\n  the fix introduce — and record it as above. Never one finding, one fix, one\n  re-review at a time.\n\nThe `verify` after a fix is often the task's third whole-suite run: past\n`tests.max_suite_runs` it exits 6 `SUITE_BUDGET`, and that run is the human's,\nfrom their own terminal, with the command it prints, or the pick in the dialog.\n\nA finding whose cause is the plan goes back to `\u002Frr-plan` for that part only. A\nreal problem outside the task is recorded with `$STATE memory risk add`; a\nfix in this task closes its risk with `$STATE memory risk close \u003Crkey> --status\nfixed --why \"…\"`. A mistake\nthis repository has seen before — the second time a review flags it — gets one\nline in the project instruction file, in this task.\n",{"name":5,"description":7,"argument-hint":24},[25,26],"B1","H1 … — a scoped re-review of these findings","main","84a91bd082ae3a289fcbe8a0f1a52bdebe5f638a","skills\u002Frr-review",0,"https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module","https:\u002F\u002Fvanssata.github.io\u002Fclaude-agentic-sdlc-module\u002F","https:\u002F\u002Fgithub.com\u002Fvanssata","2026-11-14 21:12:58",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add vanssata\u002Fclaude-agentic-sdlc-module --skill rr-review",{"tool":41,"command":42},"bunx","bunx skills add vanssata\u002Fclaude-agentic-sdlc-module --skill rr-review",{"tool":44,"command":45},"pnpm","pnpm dlx skills add vanssata\u002Fclaude-agentic-sdlc-module --skill rr-review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-review\u002FSKILL.md",true,[],[50,57,64,70,77,84],{"skillNo":51,"name":52,"nameZh":6,"description":53,"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},"01M7CF1FY7PYAK5378YCV5RV5S","sylius-fixtures","Load the Sylius fixture suite into the local database","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsetup\u002Fnew-shop-like\u002Fdot-claude\u002Fskills\u002Fsylius-fixtures\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},"01M7CF1CTV8SNSJEGJ88HFM6JC","deploy-staging","Deploy the current branch to staging and smoke-test the checkout. Use on \"deploy to staging\".","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsetup\u002Fnew-shop-like\u002Fdot-claude\u002Fskills\u002Fdeploy-staging\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"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":67,"categories":68,"keywords":69,"publishedAt":19},"01M7CF1ARVD7MZGDM7FJKE8CNR","rr-status","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-status\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7CF136MTETD8RF1NYJGYHK7","rr-spec","Turn a rough idea into one file, docs\u002Fsdlc\u002Fspecs\u002F\u003Cslug>.md — an `## Intent` section (problem, outcome, affected users and systems, constraints, decisions taken) from a short one-question-at-a-time brainstorm recorded through state.py, then requirements and design from the architect agent on the STRONG tier, checked against the project's constitution, instruction files, skills and ADRs, ending with explicitly flagged concerns (especially contradicting policies). Accepts an existing docs\u002Fsdlc\u002Fintent\u002F\u003Cslug>.md instead of the brainstorm. Use when starting a feature, \"write the spec\", \"let's define what we want\", \"design this\", before any plan or code.","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-spec\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7CF0WJVAB1ZCC0GF6NZ7H87","rr-ship","Close the task — state.py close measures the two facts (a green whole-suite verify on the tree as it is now; every changed file inside the task's files), applies the mode table and either archives the task or, in full mode, requests the human's approval and prints the exact approve command for their own terminal. Writes the commit message body and suggests the commit; never merges or deploys, and commits, pushes or approves only as the command the human picked in the dialog, verbatim, once. Use on \"\u002Frr-ship\", \"close the task\", \"ship it\", \"ready for approval\", after \u002Frr-build or \u002Frr-review.","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-ship\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"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":87,"categories":88,"keywords":89,"publishedAt":19},"01M7CF0SZAWZFAJBEK64M27KRF","rr-setup","https:\u002F\u002Fgithub.com\u002Fvanssata\u002Fclaude-agentic-sdlc-module\u002Fblob\u002Fmain\u002Fskills\u002Frr-setup\u002FSKILL.md",[],[]]