菜单
Compacts verbose context into structured summary. Use after pollution sources (searches, logs, JSON) or at phase milestones.
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
Audit and tighten codebase quality gates — lint, types, tests, code review. Use when onboarding a project, before a release, or when validation is too permissive.
Audits and aligns project documentation against its own declared hierarchy. Discovers authority chains from CONTRIBUTING.md (or equivalent), then detects broken links, duplicates, and misplaced content. Use when reviewing doc health, fixing stale references, or enforcing single-source-of-truth.
Implements concise, streamlined Python code matching exact architect specifications. Use when writing Python code, creating modules, or when the user asks to implement features in Python.
Investigates codebase before planning. Use before any non-trivial implementation task to gather context in isolation.
Provides concise, focused code reviews matching exact task complexity requirements. Use when reviewing code quality, security, or when the user asks for code review.
Writes tests following TDD (using pytest and Hypothesis) and BDD best practices. Use when writing unit tests, integration tests, or BDD scenarios.
| name | compacting-context |
| description | Compacts verbose context into structured summary. Use after pollution sources (searches, logs, JSON) or at phase milestones. |
| compatibility | Designed for Claude Code |
| metadata | {"allowed-tools":"Read, Grep, Glob","argument-hint":["compaction-name"],"context":"fork","agent":"Explore","model":"sonnet","stability":"stable","content-hash":"sha256:0a622edf25a7ca7d635e463c431ca32d74699c2908bbeca5efe7d674afae163f","last-verified-cc-version":"1.0.34"} |
Distills verbose outputs into structured summaries following ACE-FCA principles.
Per references/context-management.md:
# Compaction: {{name}}
## Trajectory
<!-- Research status → Planning status → Implementation status -->
## Key Files
<!-- files touched with purpose -->
## Completed
<!-- done items -->
## Blockers
<!-- blocking issues (empty if none) -->
## Findings
<!-- key discoveries, distilled outputs from logs/JSON/tests -->