菜单
Local code knowledge-graph + cache (no MCP, no network) — index a repo and retrieve ranked file:line for a topic instead of grepping whole files; show graph/cache status; blast-radius / dependents of a file; refresh or reset. Use for "knowledge graph status", "index this repo", "what depends on <file>", "find the code for <thing>".
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
Systematic performance profiling and optimization with a mandatory baseline-first gate — measure before changing, re-measure after. Use when something is slow or you need to hit a latency/throughput target.
Blameless post-mortem / RCA — gathers facts, reconstructs a timeline, runs 5-Whys, defines action items, writes the post-mortem file and commits it. Use after an incident or outage.
Structured design dialogue with a visual companion — produces an approved spec and design artifacts before any implementation begins. Use at the start of a feature, before writing code.
Thorough, research-grounded code review of a change, PR, or MR — multi-lens (correctness, security, simplification, tests), severity-ranked with impact + fix, configurable depth, optional PR comments. Use when asked to review code, "do a code review", "review this PR/MR", "review my changes/diff/branch", or audit a changeset before merge.
Use when the user asks to research, investigate, analyze, find out, explore, examine, audit, or evaluate something — structured multi-source research with consulting, library-doc search, web search, and guided output delivery.
Multi-agent code review AND remediation — dispatches correctness, security, and simplification reviewers in parallel, consolidates findings, then fixes criticals/highs. Use when reviewing a diff or PR before shipping.
| name | knowledge-graph |
| description | Local code knowledge-graph + cache (no MCP, no network) — index a repo and retrieve ranked file:line for a topic instead of grepping whole files; show graph/cache status; blast-radius / dependents of a file; refresh or reset. Use for "knowledge graph status", "index this repo", "what depends on <file>", "find the code for <thing>". |
Read ../../references/codex-adapter.md, then read ../../../skills/knowledge-graph/SKILL.md and follow the source skill through that Codex adapter. Keep the source skill's gates and rules.
Codex equivalents:
<plugin-root>/bin/kg path: check | init [--max N|--all] | refresh | status [--json] | query "<text>" [--k N] | neighbors <sym|file> | blast <file|sym> | stale | cache stats|clear | daemon start|stop|status | reset. Pure stdlib by default; don't hand-write graph logic.MAGICIAN_HOME to the shared adapter's Codex state root before invoking kg; never default Codex runs to ~/.claude. The graph remains global and per-repo under that root.kg init (first build on a large repo) and kg reset (destroys index+cache) are side effects: state the repo (and rough file count for a build), get explicit approval first. Reads need none.knowledge-graph: "disabled" in <magician-state>/integration-prefs.json for proactive suggestions; a direct request may update it only with write authorization.kg status --json then render via the available visualization tool. Plain-text kg status is the default.$magic uses kg query as an internal codebase source; $divine uses kg blast for change blast-radius.