Skills

Tracework ships eight independent skills. They share one storage convention, so each output can become evidence for later recall, query, brief, review, or roadmap work.

Start

Cold Start Interview

Trigger: /tracework:cold-start-interview

Runs the first-time setup for Tracework: choose a local knowledge vault, set the project identity, and save the preferences that future capture, recall, and review skills will use.

Capture

Trigger: /tracework:capture, /tracework:capture checkpoint, 收工

Captures a session or checkpoint. It dynamically routes the session to lite, standard, or deep depth, then preserves the goals, state changes, decisions, rejected paths, risks, source refs, artifact changes, and next steps worth reusing. Zero-config mode returns structured Markdown in the conversation.

Recall

Trigger: /tracework:recall, 开工

Reads recent raw entries, decision indexes, and artifact indexes to prepare the next session. It should surface decisions, risks, open questions, abandoned alternatives, and possibly stale artifacts.

Ask

Query

Trigger: /tracework:query, why did we choose this?, 为什么当时这么选

Answers targeted project-history questions from local evidence. It distinguishes provenance from verification: source_entry_refs show where something was recorded, while direct evidence refs support stronger claims. If the record is not enough, query should refuse to invent an answer.

Roadmap

Trigger: /tracework:roadmap, 决策路线图

Synthesizes decision threads, accumulating risks, recurring questions, and revisited alternatives from raw entries and decision indexes.

Review

Daily

Trigger: /tracework:daily, 日报

Writes a workplace-facing daily report from raw entries. Git history is used only to fill coverage gaps, and the output keeps project, status, risk, next step, and evidence-boundary fields stable for monthly review. Git-only output is usable but limited; capture improves the why and evidence boundary.

Weekly

Trigger: /tracework:weekly, 周报

Builds a weekly brief outline from raw entries first, with git as coverage and fallback evidence only. Report-local O#, W#, D#, and E# chains make the brief inspectable. It can run before capture coverage is complete, but fallback-only streams stay limited.

Monthly

Trigger: /tracework:monthly, 月报

Builds a monthly review from workplace daily reports and matching raw evidence. Counts and activity metrics stay in coverage context, not outcome claims.