菜单
Update the ADR inventory table in docs/adrs/README.md by scanning all adr-*.md files, extracting their title, status, and date. Use when a new ADR is added or an existing ADR's status or title has changed.
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。
基于 SOC 职业分类
Testing patterns and anti-patterns for this codebase. Use when writing tests, reviewing test coverage, or debugging test failures. Triggers on terms like "test", "assert", "coverage", "test suite", "regression".
YAML semi-indexing implementation details and debugging patterns. Use when working on YAML parsing, semi-index construction, or cursor navigation. Triggers on terms like "YAML index", "YAML parser", "sequence item", "block sequence", "YAML cursor".
Review Architecture Decision Records (ADRs) for structural accuracy, project conformance, and software engineering best practice. Works with existing project ADRs and with proposed ADRs from issues or other sources. Triggers on terms like "review ADR", "check ADR", "evaluate ADR", "ADR review".
Update the docs/ knowledge wiki when adding or changing data structures, modules, optimizations, SIMD paths, or paper references. Triggers on "knowledge map", "wiki", "update wiki", "add to wiki".
Best practices for writing and structuring Claude Code skills. Use when creating new skills, improving existing skills, or asking how to structure a SKILL.md file.
Manages benchmark documentation across multiple platforms. Use when updating benchmark results, adding performance data, or documenting jq comparison benchmarks. Triggers on terms like "benchmark", "performance", "jq comparison", "benchmark results", "update benchmarks".
| name | update-adr-inventory |
| description | Update the ADR inventory table in docs/adrs/README.md by scanning all adr-*.md files, extracting their title, status, and date. Use when a new ADR is added or an existing ADR's status or title has changed. |
| allowed-tools | Read, Edit, Glob, Grep, Bash(git log *) |
Rebuild the inventory table in docs/adrs/README.md so it reflects the current
set of ADR files and their metadata.
Only touch rows that actually changed. When adding a new ADR, append a row matching the existing column widths — do NOT reformat the header, separator, or unchanged data rows. When updating an ADR's status or title, edit only that row's cell content and pad to match the existing column width. The goal is a minimal, reviewable git diff.
Read the current inventory. Read docs/adrs/README.md and parse the
existing table to understand current column widths and which ADRs are already
listed (with their status and title).
Find all ADR files. Glob for docs/adrs/adr-*.md (exclude README.md).
Sort results in natural numeric order by ADR number.
Extract metadata from each ADR file. For every file, read the first ~10 lines and extract:
# ADR-NNNN: Title## Status section (the first non-empty line after
## Status)Determine the date. Use git to get the author date of the commit that first introduced the file:
git log --diff-filter=A --follow --format="%as" -- <file> | tail -1
If the file is untracked (no git history), use today's date.
Map status text to the table format. Normalise the raw status string (strip any existing emoji) and apply the project's status legend:
| Raw (case-insensitive) | Table display |
|---|---|
| Proposed | 🟡 Proposed |
| Accepted | ✅ Accepted |
| Deprecated | ❌ Deprecated |
| Superseded | 🔄 Superseded |
If the status does not match any of the above, use the raw text as-is.
Compare and apply minimal edits.
Only widen columns (header, separator, and all data rows) if new content exceeds the current column width. Never shrink columns.
Report what changed. Tell the user which ADRs were added, removed, or updated compared to the previous inventory, or confirm that the inventory was already up to date.