v0.3.0 — trust-layer repositioning, mission prune, engine diet, verifier hardening

Trust-layer repositioning: standalone verification, mechanical gates for the shipped surface, an instruction-budget diet, and a pruned mission catalog.

Added

• fleet-verify CLI + GitHub Action — scripts/fleet_verify.py replays every verification layer (findings re-quote, blind-fix anti-anchoring, archive manifest + sha256, fleet-outcome gates, trace schema) against one .fleet/runs/<run_id>/ directory, from this framework or any agent's artifacts. Truly-absent artifacts SKIP by name, but a missing manifest (or a manifest naming evidence a layer then can't verify) is a hard fail; an empty or manifest-less run dir exits 2 (empty verification is not success). Ships with a composite action.yml for PR-side gating and a mutation guard (fleet-verify-empty-run-not-success).
• reviewer_mode in fleet-outcome — enum (cross-vendor-structural | same-vendor-instructed | single-process-instructed) enforced by the validator: invalid values fail, absence warns. Spec in references/fleet-outcome.md; example fixture records cross-vendor-structural; mutation guard (reviewer-mode-enum-guard-off).
• Shipped-mission evidence gate — validate_mission_promotion.py --require-shipped asserts every mission in skills/ carries the progress + readiness + archive triple, resolving ledger names through the mission registry (adversarial's arch-build-* prefix included); wired into validate-all.sh with zero escapes. A documented --known-gap hatch exists but is unused.
• Trigger-loaded engine references — engine-autonomy.md, engine-workers.md, engine-review.md, engine-recovery.md under autonomous-fleet-core/references/.

Changed

• engine.md 944 → 264 lines — always-read core capped at 300 lines by check_instruction_budget.py (docs/instruction-budget.json ratchet); all doctrine moved verbatim into the four trigger-loaded references; every deep link retargeted.
• README repositioned — leads with the verification substrate, answers the single-agent-first critique, documents fleet-verify, and fixes the dead Orca link (diggerhq/orca → stablyai/orca).
• Architecture/menu counts, plugin description, and guide/docs-site catalog updated to 12 exploratory + 6 archived; test suite now 72 files / 1523 tests.

Removed

• Six exploratory missions parked to docs/exploratory/missions/archive/ — release-document, devex-audit, landing-page-convergence, product-framing, security-cso-audit, legacy-rebuild — for overlap with the shipped surface or missing enforcement machinery. Un-parking requires the evidence triple plus a written differentiation case (see the archive README). gstack-quality campaign is now an archived stub; active gstack community-preflight slugs reduce to browser-qa-fix + incident-investigate.

CI / infrastructure

• Restored main CI to green (red since 2026-06-24): fixed a ShellCheck SC2034, made the skill-creator install restore the pinned skills-lock.json, and made ~20 environment-dependent tests hermetic (fake authenticated gh + git identity shim, sandbox-availability guard, ImageMagick for banner normalization).

---

Verification: validate-all green, 100% line coverage of scripts/*.py, 72/72 mutation guards caught. Full changelog: CHANGELOG.md.

v0.2.3 — AO mechanism ports + attribution

Agent Orchestrator mechanism ports (#67) with explicit upstream attribution.

Added

• AO mechanism validators — nudge dedup (verify_nudge_dedup.py), stacked PR status (verify_stacked_pr.py), hook-signal health (verify_hook_signal.py), review supersede on sha-pin records; wired into validate-all.sh with FLEET_DISABLE_* kill switches.
• references/ao-adoptions.md — port map from Agent Orchestrator mechanisms to AF artifacts.
• ATTRIBUTIONS.md + NOTICE — credits for Agent Orchestrator (Copyright 2026 Untrivial, Apache 2.0).
• Example fixtures: .fleet/runs/example-fixture/nudge-state.json, pr-snapshot.json.
• Tests: 1368 total (new AO adoption test suites).

Changed

• engine.md AO MECHANISMS section; substrate-disable-knobs.md; Grok/Claude/Codex adapters declare activity_hooks: true.
• verify_sha_pin.py multi-file sha-pins/ layout + supersede enforcement.
• README Acknowledgments section pointing at attribution files.

Full changelog: CHANGELOG.md

v0.2.2

[0.2.2] - 2026-06-27

Substrate hardening (#64) plus warn-tier
community skill preflight and install bundles (#65).

Added

• Community dependency tiers — documented in composition.md and community-skills.md (Tier 0–4, on-missing behavior).
• community-recommends metadata on six gstack exploratory missions with mode: warn.
• community_preflight.py + preflight-community.sh — advisory checks, always-on install hints, COMMUNITY_PROBE_HOME for isolated probes.
• install-community.sh — bundle table (gstack-browser, gstack-framing, gstack-security, gstack-devex, gstack-ship, gstack); dry-run default, --execute opt-in.
• gstack-quality campaign preset — framing → qa → security → devex with allow_exploratory_nodes.
• Substrate kill-switches (FLEET_DISABLE_*) with fail-closed override semantics for security gates.
• Reviewer-role sandbox (run-sandboxed.sh --role reviewer) — macOS sandbox-exec / Linux bwrap / hash-assertion fallback.
• scripts/campaigns/audit-gated.yaml — conditional campaign with --probe-fail reachability in validate-headless.sh.
• registry_lint external pin checks — reject placeholder ref values; skill-creator pinned to anthropics/skills HEAD.
• sync_guide_starlight.py --check — fails on content drift and orphan generated docs-site pages.
• --wiring-only adapter preflight — headless/campaign dry-runs skip runtime binary checks.
• Tests: test_community_preflight.py, expanded test_reviewer_sandbox.py, test_registry_lint_cli.py, test_sync_guide_starlight.py (1312 tests).

Changed

• run-mission-headless.sh and run-campaign.sh invoke community preflight after adapter preflight.
• install-skills.sh accepts --with-community BUNDLE (opt-in); preserves Grok starter-set defaults and comments.
• setup-autonomous-fleet fleet-config template records community bundle choice.
• validate-headless.sh — gstack-quality + audit-gated presets; echoes community preflight output.
• validate-all.sh — Starlight sync uses --check (no silent working-tree mutation).
• skill_lint.py — community-recommends lint for gstack missions; lock version/hash sync retained from #64.

Fixed

• Dry-run no longer appends .fleet/runs/ to external repo .git/info/exclude (#64).
• macOS reviewer sandbox-exec profile — file-read* with scoped writes (narrow subpath profiles SIGABRT on recent darwin).
• Dogfood run-archive manifest bytes/mtime ordering for 20260626T200255Z-adversarial-review-and-fix-8358f1.
• agent-skills bundle blocks shell --execute (Claude Code slash commands only).
• skills-lock.json computedHash refreshed for autonomous-fleet-core, fleet-program, setup-autonomous-fleet.