Content
68%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-structured, token-efficient instruction skill with notably concrete wiki conventions and dated-assertion formats. Its weaknesses are the underspecified Schedules section and the absence of any explicit sequencing or validation checkpoints across the practices it prescribes.
Suggestions
Flesh out the Schedules section with one or two concrete example schedules (e.g. a weekly wiki lint, a daily progress-check cron) instead of the single generic sentence.
Add explicit sequencing or checkpoints — e.g. when to Ingest vs. Query the wiki relative to taking on a task, and a verify step before marking a strategic environment change complete.
Consider moving the detailed wiki authoring rules (dated assertions, commit citations, supersession) into a reference file and keeping a short summary in SKILL.md, since the body is slightly over the single-file threshold.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — e.g. the wiki's purpose is justified in one sentence ("只存在于你 context 里的 tribal knowledge 会在 compaction 时丢失") with no padding or explanations of known concepts. A few motivational sentences ("你有权想得比你习惯的更长远", "在改进环境这件事上不要松懈") carry no operational content, keeping it just short of a 5. | 4 / 5 |
Actionability | Mostly concrete, executable guidance: exact assertion formats ("2026-10-13，Matt 决定……"), commit-citation patterns ("在 `mattpocock/skills@abc1234` 时……"), named artifacts (`index.md`, `CODING_STANDARDS.md`), and specific practices (background agents, context pointers). The Schedules section is only two vague lines with no example schedule, which is the main gap. | 4 / 5 |
Workflow Clarity | Practices are listed (tactical/strategic tracks, subagent delegation, wiki Ingest/Query/Lint) but never sequenced — no ordering of when each behavior runs, and Lint is the only checkpoint-like step. As a role skill it has no destructive/batch operations, so no cap applies, but checkpoints remain implicit. | 3 / 5 |
Progressive Disclosure | A single self-contained file with no bundle directories, organized under clear headers (Schedules, Subagents, Wiki, 战略视角). At ~63 lines it sits just past the rubric's under-50-line simple-skill exception, and the detailed wiki sub-section ("记录历史，而不是当下") could live in a reference file, so it lands at 'good structure, minor organization gaps'. | 4 / 5 |
Total | 15 / 20 Passed |