Skip to content

Agent Context Patch v0.3.1

Choose a tag to compare

View all tags
@Cherwayway Cherwayway released this 11 Jul 15:32
Immutable release. Only release title and notes can be modified.
v0.3.1
1cf5d95
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Agent Context Patch v0.3.1

Release commit: 1cf5d95
Artifact SHA-256: ccd657ccf5784f2c0dc954dfd73210deae0d2607114e9a395028bfcdfaf99c9b

Highlights

  • Adds an experimental, on-demand personal multi-repository dogfooding Playbook without adding a public command.
  • Adds this repository's own valid Schema 1 dogfood context.
  • Adds fresh-context positive and negative acceptance cases so repeated verification failures evolve while one-off work stays out of durable context.
  • Keeps Proposal, PatchPlan, Decision, Apply Attempt, cleanup, promotion, and approval semantics delegated to the canonical v1 protocol.

Compatibility

  • Kit Version: 0.3.1.
  • Workspace Schema: reads and writes Schema 1. No workspace migration is required.
  • The update replaces only the user-level evolve skill. It does not scan or modify workspaces or AGENTS.md / CLAUDE.md.
  • The personal Playbook is experimental owner dogfooding guidance and loads only when explicitly requested.
  • Start a new Agent task after a successful update so the new skill is loaded.

Upgrade from v0.3.0

Run $evolve update, review the exact UpdatePlan and plan hash, then approve the candidate replacement. The tested path preserves a complete v0.3.0 recovery copy and leaves the workspace untouched.

Verification

  • PR #3 passed npm test on Windows and Ubuntu.
  • Exact release archive passed npm test: 139 tests, 120 passed, 19 platform skips, 0 failures on Windows.
  • The real immutable v0.3.0 asset upgraded to this exact candidate through UpdateDryRun and UpdateApply; the installed Skill reported 0.3.1, retained the v0.3.0 backup, and a second dry-run returned NoUpdate.
  • A fresh-context Agent correctly classified the positive recurrence and negative one-off acceptance cases with no lifecycle contradiction.

Current limits

The Playbook is an owner experiment, not a validated external-user default. There is no daemon, telemetry, automatic promotion, silent upgrade, or automatic workspace migration. Review the 30-day personal evidence before expanding the public command or core abstraction surface.

Assets 4
Loading