feat(skills): ship the autoresearch director protocol as public content

[aarontrowbridge]

Closes #466. Part of #390 (the content half).

Ships the director's loop protocol in the in-repo public skill library — surface: public, engine-neutral, personal-vault paths genericized — promoted from the staged deployment + armonissima #58 canonical. The protocol the studio has run all week (ledgers, trio, spec gates, CHECKOUTS, compaction-any-time safety) becomes product content.

The agent-card shipping machinery (extension prep emitting .opencode/agents/) stays with #390's own AC (role subagent registration); the armonissima internal copy retires in a follow-up vault PR once this merges.

Summary by CodeRabbit

• Documentation
  • Added public guidance for the autoresearch workflow.
  • Documented vault contracts, session tracking, recovery after context compaction, and the hypothesis-to-analysis experiment cycle.
  • Clarified subagent roles, tooling limitations, probe and experiment classification, checkout coordination, anti-gaming rules, and operational failure handling.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: a2c80125-4591-4170-9189-e21d45d7610e

📥 Commits

Reviewing files that changed from the base of the PR and between efb652d and ccc930d.

📒 Files selected for processing (1)

• packages/extension/skills/autoresearch/SKILL.md

---

📝 Walkthrough

Walkthrough

Added the public autoresearch skill protocol. It documents vault contracts, session ledgers, the research loop, agent roles, checkout coordination, compaction recovery, anti-gaming controls, and operational failure handling.

Changes

Autoresearch protocol

Layer / File(s)	Summary
Protocol contract and agent roles packages/extension/skills/autoresearch/SKILL.md	Defines public skill metadata, vault contracts, entry points, agent permissions, spec review, and probe-versus-experiment rules.
Session ledger and research execution packages/extension/skills/autoresearch/SKILL.md	Defines the session ledger, research loop, checkout claims, update triggers, and compaction recovery.
Experiment integrity and failure handling packages/extension/skills/autoresearch/SKILL.md	Defines anti-gaming controls, evidence requirements, promotion restrictions, and operational failure handling.

Estimated code review effort: 1 (Trivial) | ~5 minutes

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch amico/issue-466-autoresearch-skill

---

_{Comment @coderabbitai help to get the list of available commands.}