[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DDAES8AS0R80DH5D6TTB8S":3,"skill-same:01M7DDAES8AS0R80DH5D6TTB8S":44},{"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},"01M7DDAES8AS0R80DH5D6TTB8S","review",null,"Standalone or phase multi-model review of a Go diff, merged into one report with severities. Keywords: review, code review, multi-model review, review my changes, review the diff, critique.","girishmotwani\u002Fatom","girishmotwani","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F20937791?v=4","01M7DD9G6RCNXB9G6A33Z0GA1P","01M7DD9G7VT0Z0JHV97PSR0GN9",0,"MIT","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Fskills\u002Freview\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: review\ndescription: 'Standalone or phase multi-model review of a Go diff, merged into one report with severities. Keywords: review, code review, multi-model review, review my changes, review the diff, critique.'\n---\n\n# \u002Fatom:review\n\nReviews a diff with **two or more different models** and merges their findings into one\nreport.\n\n## Why more than one model\n\nA model reviewing its own output shares its own blind spots. It will not flag the\nmisunderstanding that produced the code, because it still holds that\nmisunderstanding.\n\n> **The reviewing model must not be the model that wrote the code.**\n\nThis is the entire point of the skill. A single-model self-review is not a substitute,\nand running one instead is the rationalization this skill exists to prevent.\n\n## When to use\n\n- For a bounded standalone review, without Atom initialization or phase state\n- After `\u002Fatom:tdd` for a phase\n- Before `\u002Fatom:ship`\n- Any time a diff is large enough that reading it once isn't enough\n\n## Writing artifacts\n\nSend complete review reports, findings, and decisions as UTF-8 standard input to\n`atom-write-artifact \u003Crelative-path>`. Treat any refusal or non-zero exit as a blocker;\nnever bypass it with a generic editor or another shell command.\n\n## Step 0 — Scope, propose and wait\n\nRead [standalone-review.md](..\u002Fatom\u002Freferences\u002Fstandalone-review.md) first.\nFollow its mode, prerequisite, consent, model-budget and artifact contract in\nboth standalone and phase contexts. Explicit review-only intent remains standalone\neven in an initialized project. Do not require phase discovery or onboarding.\n\nPropose the exact target\u002Fdiff, actual models, runs\u002Fconcurrency, evidence and file\neffects before availability probes or review submissions. Wait for approval;\nrefusal launches no reviewers and creates no bundle. Do not call `atom review`\nmerely to discover the diff or probe models before that gate.\n\n## Step 1 — Establish the diff without calling reviewers\n\n```bash\ngit diff main...HEAD             # inspect committed changes against agreed main\ngit diff --staged                # inspect staged changes\ngit diff                         # inspect unstaged working changes\n```\n\nChoose only the intended source\u002Freference; main above is an example, not a default.\nInspect surrounding code as needed for triage. The CLI supplies the selected diff\nand its evidence, not an automatic whole-file context or approved design spec.\nDo not claim reviewers received context that was not actually supplied.\n\n## Step 2 — Let the toolchain answer first\n\nNo separate command: the gate runs inside `atom review`, before the models.\n\nFor a normal review with a root Go module, Atom runs `go build`, `go vet`, `gofmt`, `go test`\nand `go test -race`, plus `staticcheck`, `errcheck` and `govulncheck` when they are\ninstalled.\n\nReview formatting checks whole surviving changed Go files from the selected\nworking-tree, staged or committed-branch snapshot, using stdin without rewriting\nsource. Untouched files are not scanned. The report names checked paths and\noutside-module or nested-module exclusions; zero applicable files means SKIPPED,\nnot a formatting pass. Other review checks retain their existing module-wide\nscope, and standalone `atom check` retains module-wide formatting.\nThose module-wide checks run against the working tree, not an isolated staged\nor committed tree. Without root `go.mod`, tooling is explicitly skipped: an exit-0\nreport does not verify a build. Missing optional tools exclude nothing.\n\n**A failing required gate normally blocks reviewer calls.** The selected-Go\ndegraded-continuation exception below must be disclosed in the proposal.\nUse explicitly approved `--strict-toolchain` when a hard stop is required.\n\nWhat passed is then written into the prompt as **excluded**: the models are told not to\nre-report an ignored error, a vet diagnostic, a formatting complaint, or a data race on\na tested path. That exclusion is the mechanism that frees attention for the rest of this\npage.\n\nThese rules keep the exclusion honest:\n\n- **A tool that did not run excludes nothing.** If `errcheck` is not installed, the\n  prompt says so and ignored errors stay in scope. Suppressing a category nobody checked\n  is how real bugs go missing.\n- **Clean is not absent.** `go test -race` only observes paths the tests actually\n  executed. Concurrent code with no test is unchecked, not safe — and the prompt says\n  that in those words.\n- **Formatting proof is file-scoped.** Only the checked snapshots are excluded\n  from formatting review; excluded modules and a skipped formatter are not clean.\n\n`--skip-toolchain` bypasses the gate. The prompt then tells the models that nothing was\nchecked and nothing is excluded, so a skipped gate costs attention rather than silently\nhiding findings.\nDo not select this bypass or another Go version as automatic failure recovery;\nthe shared entry contract requires approval of changed evidence\u002Fscope.\n\n### Optional exact Go selection\n\nThe launch module's existing `.atom\u002Fconfig.json` may set `review_go_version` to an\nexact stable `1.X.Y` patch (X >= 21). `--go-version 1.X.Y` overrides it for one run;\n`--go-executable PATH` selects an already installed direct `bin\u002Fgo` (Windows:\n`bin\u002Fgo.exe`) and requires an effective version. No installation, automatic upgrade,\nancestor lookup or version inference from `go.mod` occurs. Malformed present config\nis an explicit error. Explicit paths resolve from launch cwd before PR preparation;\ntarget checkout policy never replaces the frozen launch request.\n\nBounded installed-location discovery may need an explicit path for custom layouts.\nSelected direct Go gates and selected-root gofmt retain existing gate semantics;\npinned Makefile build\u002Ftest delegation fails before probes or recipes rather than\nbypassing the project's build. Unpinned delegation is unchanged. Advisory helper-Go\nsettings are selected, not analyzer binaries\u002Fcompiler internals or model tools.\nChild-only settings do not change defaults or make dependency activity network-free.\n\nReports\u002Fprompts distinguish requested from observed versions. Configured pins with\n`--skip-toolchain`, or requests outside a root Go module, remain `not_run` with no\nobserved executable\u002Fprobes. Explicit Go flags plus `--skip-toolchain` are usage\nerrors. Missing, mismatched or changed executable identities still block review;\ncompiler text alone cannot justify continuation.\n\nBefore approval, name the effective launch Go pin and any explicit executable.\nWithout `--strict-toolchain`, a verified selected Go older than the root module\nminimum can qualify for degraded continuation after a proven selected build\nfailure. Compatible control gates must pass and the checked input fingerprint\nmust remain unchanged before reviewer calls. This can apply to `review_go_version`\nwithout CLI Go flags; not every pin or failure qualifies. Preserve the failures,\ncontrol evidence and warning: a degraded review exits nonzero, not clean or\ngate-passed. This existing runtime policy is not permission for an agent to retry,\nswitch versions or expand the approved plan. Propose `--strict-toolchain` explicitly\nif desired; do not silently change defaults.\n\n## Step 3 — Run the models\n\n```bash\natom review --branch main --model MODEL_A --model MODEL_B --runs 1 --concurrency 1\n```\n\nRun only the complete approved command. MODEL_A and MODEL_B are placeholders for\nactual approved available models, neither the author. Use the agreed selector;\nworking\u002Fstaged alternatives are in the shared reference. Runs=1\u002Fconcurrency=1 is\na bounded recommendation, not the CLI default.\n\nEach model run receives the full selected diff and evidence, not a separate\ninvocation for each file. Runs do not receive each other's findings; concurrency\nis controlled by the approved option. The CLI uses repeated `--model` values or\nbuilt-in defaults, not general model-role resolution from `.atom\u002Fconfig.json`.\nReport unavailable, silent, unparsed or insufficient independent responses\nhonestly; do not substitute a clean review or automatically retry.\n\n## Step 4 — What reviewers look for\n\nOnly successful, applicable toolchain evidence is excluded. Unchecked concerns remain\nin scope. What is left is\nthe set of questions no tool can answer — and it is deliberately short, because a\nreviewer handed thirty concerns reports on none of them well.\n\n| Priority | Category | Why no tool catches it |\n|---|---|---|\n| 1 | **Wrong logic** | It compiles, vets clean and the tests pass — because the tests assert the wrong thing |\n| 2 | **Untested concurrency** | The race detector only observes paths the tests ran. A goroutine with no test is unchecked |\n| 3 | **Misleading names** | A name that says the wrong thing costs every future reader, and no linter can tell |\n| 4 | **Code that resists extension** | A wide interface or a package-level dependency forces the next change to edit working code |\n| 5 | **Contract drift** | The code disagrees with the approved design spec, which no tool has read |\n| 6 | **Tests that cannot fail** | A green suite that asserts nothing looks identical to a green suite that works |\n| 7 | **Unvalidated input** | `govulncheck` finds known holes in dependencies, never the one you just wrote |\n\nSeverity still comes from consequence. A finding that cannot be stated as \"this breaks\nwhen X\" is a preference, and preferences are the findings reviewers drop first. That is\nmeasured, not assumed: seeded defects phrased as style preferences were reported by\nnobody, while defects with a visible consequence were reported by everybody — even when\nthe reviewer was handed the exact rule.\n\n[go-style.md](..\u002Fatom\u002Freferences\u002Fgo-style.md) is the reference behind priorities 3 and 4.\nUse it to explain a finding, not as a checklist to walk: thirty rules handed to a\nreviewer produce worse coverage than seven.\n\nFormatting in the checked file snapshots is **not** reviewed. `gofmt` settled it\nfor those files only; a skipped formatter or an excluded module proves nothing\nabout formatting there.\n\n> **Where the real instruction lives.** This table guides *your* triage. The sentences\n> actually sent to the models are `REVIEW_PROMPT` and the toolchain block in\n> `tools\u002Fatom\u002Freview.py`. Changing this table alone changes nothing the models see —\n> if you want different findings, change the prompt and re-measure.\n\n## Step 5 — Capture the actual merged report\n\nThe runtime groups findings by file and line proximity, retaining every message\nand severity disagreement. This is not semantic deduplication: nearby reports\nmay describe different defects. It emits Markdown or JSON to stdout.\n\nPersist the report at the approved mode-specific path from the shared reference.\nStandalone uses its small review\u002Ffindings\u002Fdecisions bundle; phase review retains\n`specs\u002F{project}\u002Fphase-NN-{slug}\u002Freview.md` and existing project audit locations.\n\nEach merged finding carries:\n\n- **Severity** — critical \u002F high \u002F medium \u002F low, with disagreements retained\n- **Agreement** — participating models and per-model reporting runs\n- **Location** — `file:line`\n- **Messages** — the original findings; a concrete suggested fix is not guaranteed\n\n**Agreement is a signal, not a verdict.** Read critical\u002Fhigh findings individually,\nincluding single-model findings. Group proximity or identical synthetic replies\nare not proof of independent corroboration or correctness.\n\n## Step 6 — Triage with the developer\n\nGo through critical\u002Fhigh findings together. For each: propose a separately\napproved fix, defer to `findings.md` with a reason, or reject with a reason.\nReview-only approval does not authorize editing source.\n\n**Reject freely, silently never.** Reviewers are often wrong about a codebase they saw\nin fragments. A rejected finding with a written reason is a good outcome; a finding\nquietly dropped is a lost bug.\n\nEvery approved code fix requires appropriate verification (`atom check` for\nphase Go implementation). Every deferral and rejection is recorded.\n\n## Step 7 — Report\n\nPresent the actual report, participation\u002Ftooling limits, triage and what remains\nopen. Always persist findings\u002Fdecisions, explicitly empty when appropriate.\nStandalone stops without fixes, phase advancement or publication. Phase review\npresents its existing gate; a next-step recommendation is not permission to act.\n\n---\n\n## Done when\n\n- [ ] Exact scope, model\u002Fprobe budget and output effects approved before calls\n- [ ] Applicable gates passed, or approved missing\u002Fskipped\u002Fdegraded evidence is explicit\n- [ ] At least two **different** models reviewed, none of them the author of the code\n- [ ] Actual supplied context and response\u002Fparticipation limits stated\n- [ ] Findings grouped, severity-ranked, with messages\u002Fdisagreements retained\n- [ ] Critical\u002Fhigh findings, including single-model ones, triaged with developer\n- [ ] Any separately approved fixes verified; no review-only source edits\n- [ ] Review, findings and decisions persisted at approved locations\n\n## Next\n\nStandalone: stop at the review result.\n\nPhase review: after its gate, recommend `\u002Fatom:ship` or the next phase's\n`\u002Fatom:poc` when applicable. Do not execute it automatically.\n","# \u002Fatom:review\n\nReviews a diff with **two or more different models** and merges their findings into one\nreport.\n\n## Why more than one model\n\nA model reviewing its own output shares its own blind spots. It will not flag the\nmisunderstanding that produced the code, because it still holds that\nmisunderstanding.\n\n> **The reviewing model must not be the model that wrote the code.**\n\nThis is the entire point of the skill. A single-model self-review is not a substitute,\nand running one instead is the rationalization this skill exists to prevent.\n\n## When to use\n\n- For a bounded standalone review, without Atom initialization or phase state\n- After `\u002Fatom:tdd` for a phase\n- Before `\u002Fatom:ship`\n- Any time a diff is large enough that reading it once isn't enough\n\n## Writing artifacts\n\nSend complete review reports, findings, and decisions as UTF-8 standard input to\n`atom-write-artifact \u003Crelative-path>`. Treat any refusal or non-zero exit as a blocker;\nnever bypass it with a generic editor or another shell command.\n\n## Step 0 — Scope, propose and wait\n\nRead [standalone-review.md](..\u002Fatom\u002Freferences\u002Fstandalone-review.md) first.\nFollow its mode, prerequisite, consent, model-budget and artifact contract in\nboth standalone and phase contexts. Explicit review-only intent remains standalone\neven in an initialized project. Do not require phase discovery or onboarding.\n\nPropose the exact target\u002Fdiff, actual models, runs\u002Fconcurrency, evidence and file\neffects before availability probes or review submissions. Wait for approval;\nrefusal launches no reviewers and creates no bundle. Do not call `atom review`\nmerely to discover the diff or probe models before that gate.\n\n## Step 1 — Establish the diff without calling reviewers\n\n```bash\ngit diff main...HEAD             # inspect committed changes against agreed main\ngit diff --staged                # inspect staged changes\ngit diff                         # inspect unstaged working changes\n```\n\nChoose only the intended source\u002Freference; main above is an example, not a default.\nInspect surrounding code as needed for triage. The CLI supplies the selected diff\nand its evidence, not an automatic whole-file context or approved design spec.\nDo not claim reviewers received context that was not actually supplied.\n\n## Step 2 — Let the toolchain answer first\n\nNo separate command: the gate runs inside `atom review`, before the models.\n\nFor a normal review with a root Go module, Atom runs `go build`, `go vet`, `gofmt`, `go test`\nand `go test -race`, plus `staticcheck`, `errcheck` and `govulncheck` when they are\ninstalled.\n\nReview formatting checks whole surviving changed Go files from the selected\nworking-tree, staged or committed-branch snapshot, using stdin without rewriting\nsource. Untouched files are not scanned. The report names checked paths and\noutside-module or nested-module exclusions; zero applicable files means SKIPPED,\nnot a formatting pass. Other review checks retain their existing module-wide\nscope, and standalone `atom check` retains module-wide formatting.\nThose module-wide checks run against the working tree, not an isolated staged\nor committed tree. Without root `go.mod`, tooling is explicitly skipped: an exit-0\nreport does not verify a build. Missing optional tools exclude nothing.\n\n**A failing required gate normally blocks reviewer calls.** The selected-Go\ndegraded-continuation exception below must be disclosed in the proposal.\nUse explicitly approved `--strict-toolchain` when a hard stop is required.\n\nWhat passed is then written into the prompt as **excluded**: the models are told not to\nre-report an ignored error, a vet diagnostic, a formatting complaint, or a data race on\na tested path. That exclusion is the mechanism that frees attention for the rest of this\npage.\n\nThese rules keep the exclusion honest:\n\n- **A tool that did not run excludes nothing.** If `errcheck` is not installed, the\n  prompt says so and ignored errors stay in scope. Suppressing a category nobody checked\n  is how real bugs go missing.\n- **Clean is not absent.** `go test -race` only observes paths the tests actually\n  executed. Concurrent code with no test is unchecked, not safe — and the prompt says\n  that in those words.\n- **Formatting proof is file-scoped.** Only the checked snapshots are excluded\n  from formatting review; excluded modules and a skipped formatter are not clean.\n\n`--skip-toolchain` bypasses the gate. The prompt then tells the models that nothing was\nchecked and nothing is excluded, so a skipped gate costs attention rather than silently\nhiding findings.\nDo not select this bypass or another Go version as automatic failure recovery;\nthe shared entry contract requires approval of changed evidence\u002Fscope.\n\n### Optional exact Go selection\n\nThe launch module's existing `.atom\u002Fconfig.json` may set `review_go_version` to an\nexact stable `1.X.Y` patch (X >= 21). `--go-version 1.X.Y` overrides it for one run;\n`--go-executable PATH` selects an already installed direct `bin\u002Fgo` (Windows:\n`bin\u002Fgo.exe`) and requires an effective version. No installation, automatic upgrade,\nancestor lookup or version inference from `go.mod` occurs. Malformed present config\nis an explicit error. Explicit paths resolve from launch cwd before PR preparation;\ntarget checkout policy never replaces the frozen launch request.\n\nBounded installed-location discovery may need an explicit path for custom layouts.\nSelected direct Go gates and selected-root gofmt retain existing gate semantics;\npinned Makefile build\u002Ftest delegation fails before probes or recipes rather than\nbypassing the project's build. Unpinned delegation is unchanged. Advisory helper-Go\nsettings are selected, not analyzer binaries\u002Fcompiler internals or model tools.\nChild-only settings do not change defaults or make dependency activity network-free.\n\nReports\u002Fprompts distinguish requested from observed versions. Configured pins with\n`--skip-toolchain`, or requests outside a root Go module, remain `not_run` with no\nobserved executable\u002Fprobes. Explicit Go flags plus `--skip-toolchain` are usage\nerrors. Missing, mismatched or changed executable identities still block review;\ncompiler text alone cannot justify continuation.\n\nBefore approval, name the effective launch Go pin and any explicit executable.\nWithout `--strict-toolchain`, a verified selected Go older than the root module\nminimum can qualify for degraded continuation after a proven selected build\nfailure. Compatible control gates must pass and the checked input fingerprint\nmust remain unchanged before reviewer calls. This can apply to `review_go_version`\nwithout CLI Go flags; not every pin or failure qualifies. Preserve the failures,\ncontrol evidence and warning: a degraded review exits nonzero, not clean or\ngate-passed. This existing runtime policy is not permission for an agent to retry,\nswitch versions or expand the approved plan. Propose `--strict-toolchain` explicitly\nif desired; do not silently change defaults.\n\n## Step 3 — Run the models\n\n```bash\natom review --branch main --model MODEL_A --model MODEL_B --runs 1 --concurrency 1\n```\n\nRun only the complete approved command. MODEL_A and MODEL_B are placeholders for\nactual approved available models, neither the author. Use the agreed selector;\nworking\u002Fstaged alternatives are in the shared reference. Runs=1\u002Fconcurrency=1 is\na bounded recommendation, not the CLI default.\n\nEach model run receives the full selected diff and evidence, not a separate\ninvocation for each file. Runs do not receive each other's findings; concurrency\nis controlled by the approved option. The CLI uses repeated `--model` values or\nbuilt-in defaults, not general model-role resolution from `.atom\u002Fconfig.json`.\nReport unavailable, silent, unparsed or insufficient independent responses\nhonestly; do not substitute a clean review or automatically retry.\n\n## Step 4 — What reviewers look for\n\nOnly successful, applicable toolchain evidence is excluded. Unchecked concerns remain\nin scope. What is left is\nthe set of questions no tool can answer — and it is deliberately short, because a\nreviewer handed thirty concerns reports on none of them well.\n\n| Priority | Category | Why no tool catches it |\n|---|---|---|\n| 1 | **Wrong logic** | It compiles, vets clean and the tests pass — because the tests assert the wrong thing |\n| 2 | **Untested concurrency** | The race detector only observes paths the tests ran. A goroutine with no test is unchecked |\n| 3 | **Misleading names** | A name that says the wrong thing costs every future reader, and no linter can tell |\n| 4 | **Code that resists extension** | A wide interface or a package-level dependency forces the next change to edit working code |\n| 5 | **Contract drift** | The code disagrees with the approved design spec, which no tool has read |\n| 6 | **Tests that cannot fail** | A green suite that asserts nothing looks identical to a green suite that works |\n| 7 | **Unvalidated input** | `govulncheck` finds known holes in dependencies, never the one you just wrote |\n\nSeverity still comes from consequence. A finding that cannot be stated as \"this breaks\nwhen X\" is a preference, and preferences are the findings reviewers drop first. That is\nmeasured, not assumed: seeded defects phrased as style preferences were reported by\nnobody, while defects with a visible consequence were reported by everybody — even when\nthe reviewer was handed the exact rule.\n\n[go-style.md](..\u002Fatom\u002Freferences\u002Fgo-style.md) is the reference behind priorities 3 and 4.\nUse it to explain a finding, not as a checklist to walk: thirty rules handed to a\nreviewer produce worse coverage than seven.\n\nFormatting in the checked file snapshots is **not** reviewed. `gofmt` settled it\nfor those files only; a skipped formatter or an excluded module proves nothing\nabout formatting there.\n\n> **Where the real instruction lives.** This table guides *your* triage. The sentences\n> actually sent to the models are `REVIEW_PROMPT` and the toolchain block in\n> `tools\u002Fatom\u002Freview.py`. Changing this table alone changes nothing the models see —\n> if you want different findings, change the prompt and re-measure.\n\n## Step 5 — Capture the actual merged report\n\nThe runtime groups findings by file and line proximity, retaining every message\nand severity disagreement. This is not semantic deduplication: nearby reports\nmay describe different defects. It emits Markdown or JSON to stdout.\n\nPersist the report at the approved mode-specific path from the shared reference.\nStandalone uses its small review\u002Ffindings\u002Fdecisions bundle; phase review retains\n`specs\u002F{project}\u002Fphase-NN-{slug}\u002Freview.md` and existing project audit locations.\n\nEach merged finding carries:\n\n- **Severity** — critical \u002F high \u002F medium \u002F low, with disagreements retained\n- **Agreement** — participating models and per-model reporting runs\n- **Location** — `file:line`\n- **Messages** — the original findings; a concrete suggested fix is not guaranteed\n\n**Agreement is a signal, not a verdict.** Read critical\u002Fhigh findings individually,\nincluding single-model findings. Group proximity or identical synthetic replies\nare not proof of independent corroboration or correctness.\n\n## Step 6 — Triage with the developer\n\nGo through critical\u002Fhigh findings together. For each: propose a separately\napproved fix, defer to `findings.md` with a reason, or reject with a reason.\nReview-only approval does not authorize editing source.\n\n**Reject freely, silently never.** Reviewers are often wrong about a codebase they saw\nin fragments. A rejected finding with a written reason is a good outcome; a finding\nquietly dropped is a lost bug.\n\nEvery approved code fix requires appropriate verification (`atom check` for\nphase Go implementation). Every deferral and rejection is recorded.\n\n## Step 7 — Report\n\nPresent the actual report, participation\u002Ftooling limits, triage and what remains\nopen. Always persist findings\u002Fdecisions, explicitly empty when appropriate.\nStandalone stops without fixes, phase advancement or publication. Phase review\npresents its existing gate; a next-step recommendation is not permission to act.\n\n---\n\n## Done when\n\n- [ ] Exact scope, model\u002Fprobe budget and output effects approved before calls\n- [ ] Applicable gates passed, or approved missing\u002Fskipped\u002Fdegraded evidence is explicit\n- [ ] At least two **different** models reviewed, none of them the author of the code\n- [ ] Actual supplied context and response\u002Fparticipation limits stated\n- [ ] Findings grouped, severity-ranked, with messages\u002Fdisagreements retained\n- [ ] Critical\u002Fhigh findings, including single-model ones, triaged with developer\n- [ ] Any separately approved fixes verified; no review-only source edits\n- [ ] Review, findings and decisions persisted at approved locations\n\n## Next\n\nStandalone: stop at the review result.\n\nPhase review: after its gate, recommend `\u002Fatom:ship` or the next phase's\n`\u002Fatom:poc` when applicable. Do not execute it automatically.\n",{"name":5,"description":7},"main","5bb33cbe2c835382b5ccda840c5268321b24ac0b","skills\u002Freview","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom","https:\u002F\u002Fgithub.com\u002Fgirishmotwani","curious about all things distributed ","2026-11-13 16:16:12",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add girishmotwani\u002Fatom --skill review",{"tool":36,"command":37},"bunx","bunx skills add girishmotwani\u002Fatom --skill review",{"tool":39,"command":40},"pnpm","pnpm dlx skills add girishmotwani\u002Fatom --skill review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Fskills\u002Freview\u002FSKILL.md",true,[],[45,52,59,66,71,78],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M7DDBYEX4YKKBSQQP3HEGJC1","tdd","Implement one phase test-first in Go: red, green, refactor, against an approved design spec. Keywords: TDD, test driven, test first, implement, red green refactor, build the phase, write the code.","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Ftdd\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M7DDBT5MMFRRBZJQ01VHNDGK","spec","Write the high-level design spec for one phase, grounded in that phase POC evidence. Keywords: spec, design, design doc, architecture, API design, describe the approach.","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Fspec\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":18},"01M7DDBQ3NXQKG74B1RCHDKK9T","ship","Commit, push and open a pull request for a completed phase, with the pipeline artifacts linked. Keywords: ship, PR, pull request, commit, push, open a PR, submit.","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Fship\u002FSKILL.md",[],[],{"skillNo":67,"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":68,"categories":69,"keywords":70,"publishedAt":18},"01M7DDBJMHEDT9VPGR1GJZD8CN","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M7DDBD2G9WVYVFCTYNYFTWAE","poc","Build, run and evaluate a throwaway Go spike that answers one falsifiable question. Produces a POC Report with verbatim evidence. Keywords: POC, proof of concept, spike, prototype, prove it, experiment, try it out.","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Fpoc\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M7DDB98QNSK2WG5MMHH7S6J4","plan","Break proven work into ordered, independently shippable phases based on POC evidence. Keywords: plan, phases, break down, roadmap, decompose, milestones, sequencing.","https:\u002F\u002Fgithub.com\u002Fgirishmotwani\u002Fatom\u002Fblob\u002Fmain\u002Ftools\u002Fatom\u002Fassets\u002Fskills\u002Fplan\u002FSKILL.md",[],[]]