agentleFS
sasuketoriisasuketorii/rev_harness_public/.claude/skills/baseline-protection/SKILL.md

baseline-protection

Skill0 starsChanged 3 months agoApache-2.0View on GitHub

Ensure baseline stability across coder→reviewer cycles. Use pre-write prompt pair + pre-touch output stub + no-write-during-agent to prevent orchestrator write amplification from causing baseline stale.

The file

Files
SKILL.md

Skill: Baseline Protection

When to use

  • slice 起票時
  • coder と reviewer を順次呼ぶすべてのケース
  • 特に dirty worktree 上、fix-review loop、複数 agent 並列実行時

Why

orchestrator が coder 起動後に reviewer prompt を書くと、reviewer が live git status を取った瞬間に coder snapshot との差分が発生し、baseline stale 判定で CHANGES REQUIRED になる。

Protocol

  1. pre-write prompt pair: coder prompt と reviewer prompt を同時に orchestrator 1 ターンで書き出す。
  2. pre-touch output stubs: coder output file と reviewer output file を 0 bytes で touch する。
  3. coder 起動 -> 完了 -> reviewer 起動の間、orchestrator は一切書き込まない。
  4. reviewer 自身の output file は baseline 比較から reviewer own activity として除外することを reviewer prompt に明記する。
  5. coder 自身の output file は dirty_sha256.txt から除外する。

Automation

./.claude/commands/lib/baseline_freeze.sh を使う。

  • snapshot <evidence_dir>: git status --porcelain=v1 と SHA-256 を保存
  • pretouch <path>: output stub を事前作成
  • verify <evidence_dir> --exclude <path>: reviewer output などを除外して baseline 差分を検証

Fallback

primitive 未起動ならマニュアル運用に戻し、本 skill の 5 steps をそのまま手順書として使う。

References

  • $CLAUDE_HOME/projects/<project-slug>/memory/feedback_baseline_protection_pattern.md
  • .agent/active/prompts/reviewer_wsA_execution_s1_provenance_v2_output.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.