[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DAMF2SRF5M5QNPSHPQY6YT":3,"skill-same:01M7DAMF2SRF5M5QNPSHPQY6YT":46},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M7DAMF2SRF5M5QNPSHPQY6YT","code-review",null,"Deep review of the working branch — code rules compliance, bugs, regressions, security, soundness. Use before opening a PR, or when a change needs scrutiny beyond \u002Fcode-rules-check.","LNSD\u002Florecraft","LNSD","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F3949095?v=4","01M7DAK7ETP75C8S8PT783SWSP","Lorenzo Delgado","01M7DAK7FBVXE30743ZG437MTN",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcode-review\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: code-review\ndescription: Deep review of the working branch — code rules compliance, bugs, regressions, security, soundness. Use before opening a PR, or when a change needs scrutiny beyond \u002Fcode-rules-check.\nallowed-tools: Bash(git diff *) Bash(git status *) Bash(git merge-base *) Bash(grep *)\n---\n\n# Code Review\n\nA thorough review of the current branch, run locally. It performs `\u002Fcode-rules-check` at review depth — fanned\nout across groups of code rules — and adds what a compliance check cannot see: logic gaps, regressions, security,\nsafety, and soundness.\n\nThe subject is the uncommitted work by default (`git diff HEAD`, `git status`). For a whole branch use\n`git diff $(git merge-base HEAD main)...HEAD`. Review only what the diff touches — an unchanged file that\nbreaks a code rule is not this change's finding.\n\n## When to Use This Skill\n\n- Before opening a PR\n- After a large, risky, or long-running implementation\n- Reviewing someone else's branch locally\n- When `\u002Fcode-rules-check` is clean but the change still warrants scrutiny\n\nFor the routine \"does this follow the code rules?\" pass after finishing a piece of work, use `\u002Fcode-rules-check`\nalone — it is a fraction of the cost and is the gate in the development workflow.\n\n## Review Checklist\n\nPlease review this code change and provide feedback on:\n\n### 1. Security Concerns\n\nReview for security vulnerabilities:\n- Exposed secrets or credentials — a literal token, key or password, and anything that reaches a log line or\n  an exception message\n- Input validation and sanitization at the boundaries: a document read from disk, its YAML frontmatter, a\n  JSON Schema file, a path supplied by the caller\n- `yaml.load` on document frontmatter where `yaml.safe_load` is the only safe choice\n- A document or schema path joined from caller-supplied parts without containing it inside the corpus\n  directory, so `..\u002F` escapes the tree being checked\n- A subprocess spawned with `shell=True`, or a command line assembled by string interpolation\n\n### 2. Principles Violations\n\nThe `principle-*` code rules are the design principles, and the `\u002Fcode-rules` catalog states each one in full — work\nfrom the catalog rather than a list here, which would cover a subset and go stale on the next edit.\n\nJudge the change against all of them. Read a full principle code rule when you need to *argue* a finding: the\nexamples and the Pragmatism Caveat are what separate a violation from a deliberate, documented exception, and\na principle finding that ignores the caveat will be rejected.\n\n### 3. Potential Bugs\n\nLook for common programming errors such as:\n- Off-by-one errors, especially in line numbers and section boundaries\n- Incorrect conditionals\n- Use of the wrong variable when several of the same type are in scope\n- `min` vs `max`, `first` vs `last`, flipped ordering\n- Iterating a `dict` or `set` in an order-sensitive operation — findings reported in a different order on\n  every run are a defect, not noise\n- A mutable default argument, or a dataclass field sharing one instance across objects\n\n### 4. Unhandled Failure Paths\n\nIdentify the paths that raise where nothing locally proves they cannot:\n- Indexing or `dict[key]` on data that came from outside — parsed frontmatter, a schema file, a config file\n- `int()`, `float()` or a `datetime` parse on an untrusted string\n- A bare `except`, or an `except Exception` that swallows the failure and returns a default — a malformed\n  document must produce a finding, never a silent pass\n- `assert` used to enforce a runtime invariant; `-O` removes it\n- A resource left unreleased when the failing path skips the cleanup\n\n**Note**: this overlaps with `docs\u002Fcode\u002Ferror-handling.md`, `docs\u002Fcode\u002Ferror-types.md` and\n`docs\u002Fcode\u002Ferror-boundaries.md`.\nVerify compliance with the project's error handling standards.\n\n### 5. Backwards Compatibility\n\nVerify backwards compatibility is maintained:\n- A changed public dataclass field — a rename, a removed default, a narrowed type — breaks every caller that\n  constructs it by keyword\n- A tightened frontmatter schema or structure file that documents already in a consuming repository cannot\n  satisfy, turning a clean corpus into a wall of findings\n- A renamed check identifier, or a changed finding shape, that a caller parsing the output depends on\n- A renamed or removed public name in a package `__init__.py`\n\n### 6. Code Rule Compliance\n\nRun `\u002Fcode-rules-check`, forcing its fan-out path regardless of diff size: one agent per group of code rules,\nspawned in a single message, each applying its code rules' `## Checklist` items to the diff. That skill owns the\nprocedure — which code rules govern a change, how groups are derived, and the report format. Do not restate\nthe code rules here; they change when `docs\u002Fcode\u002F` changes.\n\nA finding in this dimension is a violation with a code rule behind it. Anything a reviewer notices that no\ncode rule states belongs in the dimensions above and below, not here.\n\n### 7. Testing\n\nEvaluate test coverage and quality:\n- Reduced test coverage without justification\n- Tests that do not actually test the intended behavior\n- Tests with race conditions or non-deterministic behavior\n- A test that mocks its own subject, or patches module internals instead of exercising the public surface\n- A document-shaped fixture inlined as a string where a checked-in file under `tests\u002F` would let the test\n  read the same thing an agent would\n- Changes to existing tests that weaken assertions\n- Changes to tests that are actually a symptom of breaking changes to user-visible behaviour\n- A new marker that is not declared in `pyproject.toml`; `--strict-markers` turns it into a collection error\n- A test that needs a network service or credentials: the suite is unit-only, so there is no tier for it and\n  `just test-unit` would either reach it or skip it silently — the dependency has to go\n\n### 8. Performance\n\nCheck for performance issues:\n- The same document read and parsed once per check instead of once per document\n- A regex compiled, or a JSON Schema loaded, inside a per-document loop rather than once up front\n- A quadratic scan over sections or findings where a single pass, or a lookup by key, would do\n- Reading and parsing a whole document when only its frontmatter is needed\n\n### 9. Documentation\n\nEnsure documentation is up-to-date:\n- `README.md` and `docs\u002F` reflect current behaviour, and the `justfile` recipes they quote still exist\n- `AGENTS.md` marks planned things as planned: a section that calls something \"planned, not created\" must be\n  updated in the same change that creates it\n- A new or changed check is reflected in the spec it enforces, under `docs\u002F__meta__\u002F` — the\n  requirement a check applies and the prose stating it change together\n\n### 10. Dead Code\n\nFind dead code that is not caught by ruff:\n- A value assigned then overwritten before it is read\n- A module, class or function nothing imports, and that no registry discovers by name\n- `breakpoint()`, a debugging `print()`, or a commented-out block left behind\n- A `TODO` or `FIXME` added by this change without an owner or an issue\n\n### 11. Inconsistencies\n\nLook for inconsistencies between comments and code:\n- A docstring whose `Args`, `Returns` or `Raises` no longer matches the signature or the body\n- Misleading variable names or comments\n- Outdated comments after refactoring\n\n### 12. Code Rule Validation\n\nIf the change touches `docs\u002Fcode\u002F` or `docs\u002F__meta__\u002F`, check the code rules against the specs that govern\nthem: the corpus spec `docs\u002F__meta__\u002Fcode.md`, narrowed by the `code-\u003Cprefix>.md` namespace spec for\nthe code rule's prefix. No checker is wired up, so this is done by reading.\n\n## Notes\n\n### Focus on Actionable Feedback\n\n- Provide specific, actionable feedback on actual lines of code\n- Avoid general comments without code references\n- Reference specific file paths and line numbers\n- Suggest concrete improvements\n\n### Code Rule Compliance is Critical\n\nCode rule violations should be treated seriously as they:\n- Reduce codebase consistency\n- Make maintenance harder\n- May introduce security vulnerabilities, in the code that parses documents from outside this repository\n  above all\n- Conflict with established architectural decisions\n\nAlways run the code rule compliance review (section 6) as part of every code review.\n\n### Review Priority\n\nSections are ordered by priority — review from top to bottom:\n1. **Security concerns** (§1, highest priority)\n2. **Principles violations** (§2)\n3. **Potential bugs** and **unhandled failure paths** (§3–4)\n4. **Backwards compatibility** (§5)\n5. **Code rule violations** (§6)\n6. **Testing** (§7)\n7. **Performance** (§8)\n8. **Documentation**, **dead code**, and **inconsistencies** (§9–11)\n\n## Next Steps\n\nAfter completing the code review:\n1. Provide clear, prioritized feedback\n2. Distinguish between blocking issues (bugs, security) and suggestions (style, performance)\n3. Reference specific patterns from `docs\u002Fcode\u002F` when flagging violations\n4. Suggest using `\u002Fcode-format`, `\u002Fcode-check`, and `\u002Fcode-test` to validate fixes\n","# Code Review\n\nA thorough review of the current branch, run locally. It performs `\u002Fcode-rules-check` at review depth — fanned\nout across groups of code rules — and adds what a compliance check cannot see: logic gaps, regressions, security,\nsafety, and soundness.\n\nThe subject is the uncommitted work by default (`git diff HEAD`, `git status`). For a whole branch use\n`git diff $(git merge-base HEAD main)...HEAD`. Review only what the diff touches — an unchanged file that\nbreaks a code rule is not this change's finding.\n\n## When to Use This Skill\n\n- Before opening a PR\n- After a large, risky, or long-running implementation\n- Reviewing someone else's branch locally\n- When `\u002Fcode-rules-check` is clean but the change still warrants scrutiny\n\nFor the routine \"does this follow the code rules?\" pass after finishing a piece of work, use `\u002Fcode-rules-check`\nalone — it is a fraction of the cost and is the gate in the development workflow.\n\n## Review Checklist\n\nPlease review this code change and provide feedback on:\n\n### 1. Security Concerns\n\nReview for security vulnerabilities:\n- Exposed secrets or credentials — a literal token, key or password, and anything that reaches a log line or\n  an exception message\n- Input validation and sanitization at the boundaries: a document read from disk, its YAML frontmatter, a\n  JSON Schema file, a path supplied by the caller\n- `yaml.load` on document frontmatter where `yaml.safe_load` is the only safe choice\n- A document or schema path joined from caller-supplied parts without containing it inside the corpus\n  directory, so `..\u002F` escapes the tree being checked\n- A subprocess spawned with `shell=True`, or a command line assembled by string interpolation\n\n### 2. Principles Violations\n\nThe `principle-*` code rules are the design principles, and the `\u002Fcode-rules` catalog states each one in full — work\nfrom the catalog rather than a list here, which would cover a subset and go stale on the next edit.\n\nJudge the change against all of them. Read a full principle code rule when you need to *argue* a finding: the\nexamples and the Pragmatism Caveat are what separate a violation from a deliberate, documented exception, and\na principle finding that ignores the caveat will be rejected.\n\n### 3. Potential Bugs\n\nLook for common programming errors such as:\n- Off-by-one errors, especially in line numbers and section boundaries\n- Incorrect conditionals\n- Use of the wrong variable when several of the same type are in scope\n- `min` vs `max`, `first` vs `last`, flipped ordering\n- Iterating a `dict` or `set` in an order-sensitive operation — findings reported in a different order on\n  every run are a defect, not noise\n- A mutable default argument, or a dataclass field sharing one instance across objects\n\n### 4. Unhandled Failure Paths\n\nIdentify the paths that raise where nothing locally proves they cannot:\n- Indexing or `dict[key]` on data that came from outside — parsed frontmatter, a schema file, a config file\n- `int()`, `float()` or a `datetime` parse on an untrusted string\n- A bare `except`, or an `except Exception` that swallows the failure and returns a default — a malformed\n  document must produce a finding, never a silent pass\n- `assert` used to enforce a runtime invariant; `-O` removes it\n- A resource left unreleased when the failing path skips the cleanup\n\n**Note**: this overlaps with `docs\u002Fcode\u002Ferror-handling.md`, `docs\u002Fcode\u002Ferror-types.md` and\n`docs\u002Fcode\u002Ferror-boundaries.md`.\nVerify compliance with the project's error handling standards.\n\n### 5. Backwards Compatibility\n\nVerify backwards compatibility is maintained:\n- A changed public dataclass field — a rename, a removed default, a narrowed type — breaks every caller that\n  constructs it by keyword\n- A tightened frontmatter schema or structure file that documents already in a consuming repository cannot\n  satisfy, turning a clean corpus into a wall of findings\n- A renamed check identifier, or a changed finding shape, that a caller parsing the output depends on\n- A renamed or removed public name in a package `__init__.py`\n\n### 6. Code Rule Compliance\n\nRun `\u002Fcode-rules-check`, forcing its fan-out path regardless of diff size: one agent per group of code rules,\nspawned in a single message, each applying its code rules' `## Checklist` items to the diff. That skill owns the\nprocedure — which code rules govern a change, how groups are derived, and the report format. Do not restate\nthe code rules here; they change when `docs\u002Fcode\u002F` changes.\n\nA finding in this dimension is a violation with a code rule behind it. Anything a reviewer notices that no\ncode rule states belongs in the dimensions above and below, not here.\n\n### 7. Testing\n\nEvaluate test coverage and quality:\n- Reduced test coverage without justification\n- Tests that do not actually test the intended behavior\n- Tests with race conditions or non-deterministic behavior\n- A test that mocks its own subject, or patches module internals instead of exercising the public surface\n- A document-shaped fixture inlined as a string where a checked-in file under `tests\u002F` would let the test\n  read the same thing an agent would\n- Changes to existing tests that weaken assertions\n- Changes to tests that are actually a symptom of breaking changes to user-visible behaviour\n- A new marker that is not declared in `pyproject.toml`; `--strict-markers` turns it into a collection error\n- A test that needs a network service or credentials: the suite is unit-only, so there is no tier for it and\n  `just test-unit` would either reach it or skip it silently — the dependency has to go\n\n### 8. Performance\n\nCheck for performance issues:\n- The same document read and parsed once per check instead of once per document\n- A regex compiled, or a JSON Schema loaded, inside a per-document loop rather than once up front\n- A quadratic scan over sections or findings where a single pass, or a lookup by key, would do\n- Reading and parsing a whole document when only its frontmatter is needed\n\n### 9. Documentation\n\nEnsure documentation is up-to-date:\n- `README.md` and `docs\u002F` reflect current behaviour, and the `justfile` recipes they quote still exist\n- `AGENTS.md` marks planned things as planned: a section that calls something \"planned, not created\" must be\n  updated in the same change that creates it\n- A new or changed check is reflected in the spec it enforces, under `docs\u002F__meta__\u002F` — the\n  requirement a check applies and the prose stating it change together\n\n### 10. Dead Code\n\nFind dead code that is not caught by ruff:\n- A value assigned then overwritten before it is read\n- A module, class or function nothing imports, and that no registry discovers by name\n- `breakpoint()`, a debugging `print()`, or a commented-out block left behind\n- A `TODO` or `FIXME` added by this change without an owner or an issue\n\n### 11. Inconsistencies\n\nLook for inconsistencies between comments and code:\n- A docstring whose `Args`, `Returns` or `Raises` no longer matches the signature or the body\n- Misleading variable names or comments\n- Outdated comments after refactoring\n\n### 12. Code Rule Validation\n\nIf the change touches `docs\u002Fcode\u002F` or `docs\u002F__meta__\u002F`, check the code rules against the specs that govern\nthem: the corpus spec `docs\u002F__meta__\u002Fcode.md`, narrowed by the `code-\u003Cprefix>.md` namespace spec for\nthe code rule's prefix. No checker is wired up, so this is done by reading.\n\n## Notes\n\n### Focus on Actionable Feedback\n\n- Provide specific, actionable feedback on actual lines of code\n- Avoid general comments without code references\n- Reference specific file paths and line numbers\n- Suggest concrete improvements\n\n### Code Rule Compliance is Critical\n\nCode rule violations should be treated seriously as they:\n- Reduce codebase consistency\n- Make maintenance harder\n- May introduce security vulnerabilities, in the code that parses documents from outside this repository\n  above all\n- Conflict with established architectural decisions\n\nAlways run the code rule compliance review (section 6) as part of every code review.\n\n### Review Priority\n\nSections are ordered by priority — review from top to bottom:\n1. **Security concerns** (§1, highest priority)\n2. **Principles violations** (§2)\n3. **Potential bugs** and **unhandled failure paths** (§3–4)\n4. **Backwards compatibility** (§5)\n5. **Code rule violations** (§6)\n6. **Testing** (§7)\n7. **Performance** (§8)\n8. **Documentation**, **dead code**, and **inconsistencies** (§9–11)\n\n## Next Steps\n\nAfter completing the code review:\n1. Provide clear, prioritized feedback\n2. Distinguish between blocking issues (bugs, security) and suggestions (style, performance)\n3. Reference specific patterns from `docs\u002Fcode\u002F` when flagging violations\n4. Suggest using `\u002Fcode-format`, `\u002Fcode-check`, and `\u002Fcode-test` to validate fixes\n",{"name":5,"description":7,"allowed-tools":24},"Bash(git diff *) Bash(git status *) Bash(git merge-base *) Bash(grep *)","main","d2f3a5835a8673ccf134ae62f0a5e84104ff7098",".agents\u002Fskills\u002Fcode-review",1,"https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft","https:\u002F\u002Fgithub.com\u002FLNSD","2026-11-10 04:14:00",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add LNSD\u002Florecraft --skill code-review",{"tool":38,"command":39},"bunx","bunx skills add LNSD\u002Florecraft --skill code-review",{"tool":41,"command":42},"pnpm","pnpm dlx skills add LNSD\u002Florecraft --skill code-review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcode-review\u002FSKILL.md",true,[],[47,54,59,64,69,74],{"skillNo":48,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M7DASA0R4R68AVTZJFJJBYDB","review","Review a change. Use when asked to review.","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC504\u002Ftrigger\u002Freference-link-to-parent-skill-file\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":55,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M7DAS1YXJQ8FGABR8BGX9FAT","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC504\u002Fnear_miss\u002Fclimb-that-stays-inside-and-document-climb\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":60,"name":49,"nameZh":6,"description":50,"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},"01M7DAS0TCKQFF2YJ5VY9SWHV2","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC503\u002Ftrigger\u002Fskill-reference-link-read-from-skill-root\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":65,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M7DARZNA9QGXHAY17142R8YT","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC503\u002Fnear_miss\u002Fskill-link-climbing-above-skill-root\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":70,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7DARYHB7WBR44AJXFJSZQ7X","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC502\u002Ftrigger\u002Fskill-fragment-in-file-without-headings\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":75,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":19},"01M7DARXCTFKMH92FHVV8H30MM","https:\u002F\u002Fgithub.com\u002FLNSD\u002Florecraft\u002Fblob\u002Fmain\u002Ftests\u002Fit\u002Frule_cases\u002FLC501\u002Ftrigger\u002Fskill-link-starting-with-slash\u002Froot\u002F.agents\u002Fskills\u002Freview\u002FSKILL.md",[],[]]