8000 docs: v0.18.0 megaplan — agent autonomy sprint (Tracks 148–158) by stpinkie · Pull Request #60 · stpinkie/rhizome · GitHub
[go: up one dir, main page]

Skip to content

docs: v0.18.0 megaplan — agent autonomy sprint (Tracks 148–158) - #60

Merged
stpinkie merged 1 commit into
mainfrom
devin/v0.18.0-megaplan
Sep 29, 2026
Merged

stpinkie merged 1 commit into
mainfrom
devin/v0.18.0-megaplan

Conversation

@stpinkie
Copy link
Copy Markdown
Owner

Summary

  • docs/design/v0.18.0-sprint.md — design record for the Agent Autonomy sprint: pkg/evolution opened to the operator review-first.
  • .todo.md — Planned — v0.18.0 section (Tracks 148–158); queue rotated to v0.19.0+.
  • ROADMAP.md — lookahead extended to four sprints; §4 gains the evolution-depth line + AIEOS cross-ref.

Track shape

Track Scope
148 Upstream cherry-pick sweep + re-triage (standing opener)
149 rhizome evolution verbs + daemon API — the documented manual-trigger gap (status/run/drafts/draft show+accept+reject/records/skills; reads daemonless, mutations via daemon)
150 Draft-review hardening — pkg/guard + pkg/redact scans → structured guard:/secret: findings; auto-quarantine policy; accept --force is the logged override
151 Rollback + version UX — rides existing backup machinery; writes the dormant SkillVersionEntry.Rollback fields; skills history
152 Apply-mode policy + provenance — apply_require_review (default true), apply_max_per_day, protected_skills, .evolution-origin.json
153 Lifecycle + retention — archive/restore verbs, bounded record/draft retention, evolution prune
154 Feedback-signal records — RecordKindFeedback joins judge evidence, never clusters (flex)
155 Observability — evolution.* events + evolution-audit.jsonl + status depth
156 Evolved-skill sharing — .evolution-origin.json rides the pull bundle; agentmanifest gains emit-when-set evolved_skills attestation
157 Web UI evolution panel (declared cut line — API lands regardless)
158 Docs & release

Key decisions

  • Review-first posture: apply_require_review defaults true — apply-mode drafts park as candidates until draft accept; =false reproduces pre-sprint auto-apply exactly.
  • Scanner upgrade reuses pkg/guard (the v0.7.1 injection scanner the pull path already trusts) rather than a third scanner.
  • Reads daemonless, mutations daemon-required — the economy-ledger ownership rule from v0.17.0 applies verbatim to evolution state.
  • Provenance is a marker, not laundering — evolved skills carry .evolution-origin.json end-to-end through mesh sharing; peers see generated:true.
  • Cut lines: 157 → 154 → 153-verbs → 156 → 152-protected_skills. Never cut: 148, 149, 150, 151, 155, 158.

Test plan

  • bash ./scripts/lint-docs.sh — OK
  • Docs-only diff; no code touched

Generated with Devin

pkg/evolution shipped as a closed loop with no operator surface, a
narrow draft scanner, and dormant rollback fields. This plan opens the
loop review-first: rhizome evolution verbs + daemon API (the documented
manual-trigger gap), guard/redact draft-review hardening, rollback UX,
apply-mode policy gates (require_review default), feedback records,
observability, and provenance-marked evolved-skill mesh sharing.

Generated with [Devin](https://devin.ai)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
@github-actions github-actions Bot added the docs label Sep 29, 2026
@stpinkie
stpinkie merged commit 0e3ec00 into main Sep 29, 2026
14 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

0