メニュー
Use when generating a human-readable onboarding document from HARNESS.md, AGENTS.md, and REFLECTION_LOG.md — produces a friendly guide for new team members joining a harnessed project
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/Habitat-Thinking/ai-literacy-superpowers --skill harness-onboardingコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
Use when acting as the adversarial spec reviewer — raises steel-manned objections across six categories before plan approval, requires evidence per objection, and discloses what was not challenged
Use when designing, reviewing, or classifying an agent that guards the human — defines the sentinel category, the three-part signature (S1 read-only, S2 advisory-to-human, S3 explicit honesty rule), the near-miss gallery that explains why read-only-plus-advisory is not sufficient, the honesty-rule-before-detection-logic discipline, and the three anti-patterns that eject an agent from the category
This skill should be used when the user asks to "assess AI literacy", "run an assessment", "check literacy level", "evaluate our AI collaboration", "where are we on the framework", or wants to determine their team's AI literacy level using the ALCI instrument.
| name | harness-onboarding |
| description | Use when generating a human-readable onboarding document from HARNESS.md, AGENTS.md, and REFLECTION_LOG.md — produces a friendly guide for new team members joining a harnessed project |
HARNESS.md is written for agents and experienced team members. New contributors need something different — a friendly, practical guide that explains "here's how we work" without requiring them to parse constraint definitions and GC rule syntax.
This skill generates ONBOARDING.md from three sources the project
already maintains. No new data entry is needed — the onboarding
document is a human-readable projection of existing harness state.
| Source | What it contributes |
|---|---|
| HARNESS.md | Stack, conventions, constraints, GC rules, observability cadence |
| AGENTS.md | Gotchas, architecture decisions, test strategy, design decisions |
| REFLECTION_LOG.md | Recent surprises with context or workflow signal |
HARNESS.md, AGENTS.md, and REFLECTION_LOG.md are the single source of truth. Sync is one-way only — these files drive ONBOARDING.md. The generated document is never edited directly.
Write for a human who just joined the team and needs to ship their first PR. The reader is a skilled developer who knows nothing about this specific project.
One paragraph. What this project is, why it matters, what a new contributor should expect. Pull from the README if it has a good description; otherwise synthesise from the stack and conventions.
From HARNESS.md Context > Stack. List each technology with a brief note on how it's used. Don't just repeat the HARNESS.md bullets — add context about what each tool does in this project.
From HARNESS.md Context > Conventions. Transform the terse bullet format into readable prose. Explain not just the rule but why the team follows it. Use examples where helpful.
From HARNESS.md Constraints. Group by when they fire:
For each constraint, explain in plain English what it checks and what happens if it fails. Skip the Tool and Enforcement fields — those are implementation details for the harness, not for the contributor.
From AGENTS.md GOTCHAS + recent REFLECTION_LOG entries where
Signal is context or workflow. Present as practical warnings:
"what happened" and "how to avoid it". Limit to the 5-7 most
relevant entries — the onboarding doc should not be exhaustive.
From AGENTS.md ARCH_DECISIONS. Each entry: what was decided, why, and what the alternatives were. Frame as "the team already decided this — here's why."
From AGENTS.md TEST_STRATEGY. Explain how quality is assured. What tools run, what they check, how to run them locally.
Brief explanation of the three enforcement loops. One sentence each for advisory (hooks), strict (CI), and investigative (GC). Mention the observability cadence from HARNESS.md Observability. Don't explain the framework — just explain what the contributor will encounter.
Synthesise from Constraints and Conventions. Include only items the contributor needs to check before pushing. Format as a numbered list they can work through. Example items:
Links to HARNESS.md, AGENTS.md, REFLECTION_LOG.md, and the docs site if one exists.
The generated document must match the section structure of
templates/ONBOARDING.md. The /harness-onboarding command
includes a validation checkpoint that verifies all 10 sections
are present and fixes missing sections in place.
A GC rule checks monthly whether ONBOARDING.md is stale — whether
any of its three sources (HARNESS.md, AGENTS.md, REFLECTION_LOG.md)
have been modified more recently than ONBOARDING.md. When stale,
the GC rule flags it and suggests re-running /harness-onboarding.