菜单
Provides review-workflow scaffolding for context, evidence, and output. Use at the start of any detailed review to ensure consistent, comparable findings.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/athola/claude-night-market --skill review-core命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Turn hunches into accepted results: worthiness score, evidence bar, research-to-rules. Use when vetting ideas. Not for QA; use night-market-validation-and-qa.
Generates and self-executes a diff-derived test plan for a PR. Use when validating PR changes before merge. Do not use for code review; use sanctum:pr-review.
Curate the web-capture index. Use when the capture backlog grows, captures sit unprocessed at seedling/pending, or to surface stored research during work.
基于 SOC 职业分类
正在显示 SKILL.md
| name | review-core |
| description | Provides review-workflow scaffolding for context, evidence, and output. Use at the start of any detailed review to ensure consistent, comparable findings. |
| alwaysApply | false |
| category | review-patterns |
| tags | ["workflow","scaffolding","evidence","reporting","analysis"] |
| dependencies | [] |
| tools | [] |
| usage_patterns | ["review-preflight","workflow-scaffolding","evidence-capture"] |
| complexity | intermediate |
| model_hint | standard |
| estimated_tokens | 1500 |
Trigger Keywords: review, audit, analysis, assessment, evaluation, inspection Contextual Cues:
Auto-Load When: Any review-specific workflow is detected or when analysis methodologies are requested.
review-core:context-establishedreview-core:scope-inventoriedreview-core:evidence-capturedreview-core:deliverables-structuredreview-core:findings-verifiedreview-core:contingencies-documentedreview-core:context-established)pwd, repo, branch, and upstream base (e.g., git status -sb, git rev-parse --abbrev-ref HEAD).review-core:scope-inventoried)rg --files -g '*.mk', ls docs, cargo metadata).review-core:evidence-captured)git diff --stat, make -pn, cargo doc, web.run citations). Keep snippets or line numbers for later reference.If this work involved rework, a failed approach, or a blocker, record it to
docs/lessons-learned.md so the insight survives past the session (draft and
confirm):
Skill(leyline:decision-journal) and append
a lesson entry (what_happened, what_didnt_work, root_cause, action;
set phase to review). Show the draft; append on confirmation.docs/lessons-learned.md using the
in-file ENTRY TEMPLATE; assign the next LL-NNN id.review-core:deliverables-structured)review-core:findings-verified)Every finding must be falsifiable: a citation a second pass can mechanically re-read and confirm. Findings that fail verification do not ship.
Use the grounded-finding schema from Skill(imbue:structured-output):
each finding carries a Location (file:line) and a verbatim
Anchor snippet copied from that line.
Write the findings to .review/findings.json (one object per finding:
id, file, line, anchor, severity, category,
recommendation, evidence_refs).
Run the verifier:
python plugins/imbue/scripts/citation_verifier.py \
--findings .review/findings.json --repo-root .
Exit 0 means every citation resolved; exit 1 lists each finding
whose path, line, or anchor did not match the source.
Drop or label UNVERIFIED any finding the verifier failed; only
verified findings enter the report. Attach the verifier output to the
evidence appendix.
If the script is unavailable, fall back to re-reading each cited
file:line by hand and confirming the anchor text is present; note the
manual fallback in the contingency section.
review-core:contingencies-documented)web.run), document the alternative steps that will be taken and any limitations this introduces. This helps reviewers understand any gaps in coverage.Location + verbatim Anchor and was confirmed by citation_verifier.py (or a documented manual re-read); no unverified findings ship..review/findings.json exists and the verifier exited 0, or every failed finding was dropped or labeled UNVERIFIED.docs/lessons-learned.md (or the in-file template).