Battam1111/Myco

v0.6.16 — 2026-04-29 — Neat-freak sweep + autopoietic-loop IOU split

Zero R1-R7 surface deltas; zero new manifest verbs; zero new lint dims; zero subsystem changes; schema v2 unchanged. This is a hygiene molt: 27-patch deterministic sweep of stale narrative refs accumulated through the v0.6.0 → v0.6.15 minor sequence (5 versions in 2 days), plus one structural decision — SPLITTING the originally-scoped v0.6.16 (which bundled autopoietic-loop completion: helper + senesce reaper + auto_merge.yml).

Why this release (owner observation)

go. 同时结合之前吃的洁癖 skill, 将整个 Myco 的项目目录结构、所有文件进行重构优化, 尽最大可能去冗余化, 使得 Myco 再一次蜕变.

Translation: ship v0.6.16 with maximum de-redundantization across the entire project, completing the substrate's "molting once again" from v0.6.15.

The 5-critic L0-P1-P5 fanout's first dogfood

The v0.6.15-shipped 5-critic L0-P1-P5 fanout pattern got its first substantive load test on the originally-scoped v0.6.16 (helper + senesce reaper + auto_merge.yml + 27-patch sweep). It surfaced 3 P0 BLOCK + 4 HIGH structural tensions in 8 minutes of parallel scrutiny:

• mycorrhiza T1 (P0) — governance.public_window_min_senesce_count: 7 requires senesce_count persistence across sessions; source-grep returns zero matches for senesce_count/session_count/sessions.db in src/myco/. Auto-LAND has no data source.
• mycorrhiza T2 (P0) — pyyaml.safe_dump cannot reproduce _canon.yaml's 6-space block indent. Block-level rebuild via pyyaml mutates visual style → ruff/format/diff noise.
• mycorrhiza T3 (P0) — gh pr merge --auto requires repo Settings → "Allow auto-merge" + Branch Protection rule with required status checks. Repo audit returns zero ruleset files.
• chytrid T3 — bundling helper + senesce + auto_merge + 27-patch sweep + 4 doctrine rewrites violates L0 P4 cadence. The v0.6.15 endophyte T9 explicitly warned: "more gates = safer is the wrong instinct; iterate small."
• mycoparasite T1, T2 — auto_merge race conditions (immediate merge after senesce LANDED; cron */6 vs vetoed_at write race).
• saprotroph T1 — canon_schema.md has zero documentation of governance.* shape; helper writing rich entries without schema docs makes L1↔L4 drift permanent.
• saprotroph T3, T4 — proposed helper in core/ violates PA4 (mechanical, HIGH); v0.6.16 sweeping L1 canon_schema.md auto-classifies as HIGH-risk.

The split decision

The 5-critic fanout's converging recommendation: SPLIT. The autopoietic-loop completion has infrastructure-level prerequisites that don't exist yet (senesce_count persistence backend, ruamel.yaml round-trip choice, Branch Protection ruleset). Bundling them into v0.6.16 forces a "design-and-ship within one molt" that violates L0 P4 (永恒迭代 — small batches, high cadence) and risks shipping a structurally-broken auto-LAND.

| version | scope | risk profile | infrastructure prerequisites |
|---|---|---|---|
| v0.6.16 (this) | Neat-freak sweep alone (~−210 LoC across 17 files) + risk_classifier dead-pattern fix | LOW for source patches; HIGH-but-owner-gated for L1 canon_schema.md sweep | None — pure SE2 corrections |
| v0.6.17 (planned) | Canon round-trip helper (cycle/governance.py, NOT core/) + senesce_count persistence backend + senesce reaper canon-direct-write | LOW (helper, infra) | Helper API design lands in this release's craft |
| v0.6.18 (planned) | .github/workflows/auto_merge.yml + 24h grace window + Branch Protection ruleset | MEDIUM | Operator step (Branch Protection settings) must precede merge |

This sequence honors L0 P4 cadence (3 small batches, not 1 big one), L0 P1 Agent-First (each batch has a clear LOW/MEDIUM risk tier; no silent owner-gate inversions), and L0 P3 perpetual evolution (each step measurable; later steps refine based on data from earlier).

What changed

Group A — Source code dead-pattern fixes (contract-grade)

• src/myco/core/risk_classifier.py:141-143 — DEAD PATTERNS in the just-shipped v0.6.15 medium-risk classifier removed. The v0.6.15 craft added recursion-cutter + HIGH-risk patterns at the new boundary paths (src/myco/boundary/surface/manifest.yaml, _canon_lint.yaml) but never updated the legacy MEDIUM-tier patterns at the pre-v0.6.0 paths (src/myco/surface/manifest.yaml, src/myco/symbionts/.*\.py). Lines 141, 142 were redundant with HIGH at lines 100, 101 (unreachable code). Line 143's symbionts/ path was excreted at v0.6.0. Cleanup: removed lines 141-142 entirely (redundancy), updated line 143 to boundary/host_integration/. This was a quietly-broken v0.6.15 ship — the recursion-cutter craft itself shipped with a stale sibling pattern set.
• src/myco/cycle/ramify.py:5 — docstring path corrected.

Group B — _canon.yaml SSoT alignment

• metrics.test_count — 1477 → 1545 (68-test drift accumulated since v0.6.0).
• waves.current — 27 → 28 (auto-incremented by myco molt).

Group C — Doctrine path-corrections (L0/L1/L2)

All references to deleted src/myco/symbionts/ package replaced with src/myco/boundary/host_integration/ (the v0.6.0 unification target). All references to deleted docs/architecture/L3_IMPLEMENTATION/symbiont_protocol.md replaced with docs/architecture/L2_DOCTRINE/boundary.md (the canonical home for boundary subsystem doctrine).

• docs/architecture/L0_VISION.md:131-136 — per-host extension axis path corrected.
• docs/architecture/L1_CONTRACT/protocol.md:131-136 — same.
• docs/architecture/L2_DOCTRINE/extensibility.md — bulk path correction across §"The two axes" + §"Per-host" + §"What neither axis is" + cross-reference matrix; v0.5.6 "defined-but-empty" claim updated to "10 of 14 hosts populated at v0.6.15"; providers/ excretion documented (excreted at v0.6.14 after seven minor releases without population).
• docs/architecture/L2_DOCTRINE/homeostasis.md:65-67 — path + doc reference corrected; v0.5.8 25-dim narrative gets a v0.6.0 46-dim refresh pointer.
• 13 src/myco/boundary/host_integration/*.py + 3 src/myco/boundary/install/*.py — docstring code_doc_ref paths corrected (cleared 13 SE1 medium findings introduced by the symbiont_protocol.md excretion).

Group D — L1 schema documentation (saprotroph T1 fold-in)

• docs/architecture/L1_CONTRACT/canon_schema.md — comprehensive refresh:
  • Top example block: contract_version/synced_contract_version v0.5.7 → v0.6.15; schema_version "1" → "2".
  • lint.dimensions block replaced with lint.dimensions_ref: _canon_lint.yaml (matches actual v0.6.0 SSoT) + 46-dim representative slice as comment.
  • subsystems block adds cycle: (6th, v0.6.0+) + boundary: (7th, v0.6.0+) entries.
  • commands.manifest_ref path corrected to src/myco/boundary/surface/manifest.yaml.
  • NEW section: "v0.6.0+ schema v2 additions" — documents system.llm_policy (v0.6.14 enum-narrowed), system.resource_redaction, system.resource_watch, and system.governance.* shapes including the last_winnowed_proposals[] object schema. Closes the saprotroph T1 finding ("zero documentation of governance.* shape would make L1↔L4 drift permanent once helper writes rich entries").

Group E — L3 implementation doctrine

• docs/architecture/L3_IMPLEMENTATION/symbiont_protocol.md — EXCRETED (186 lines). The page was anchored to the deleted src/myco/symbionts/ package; superseded by L2_DOCTRINE/boundary.md "Sixth seam" + "Subagents and slash commands" sections.
• docs/architecture/L3_IMPLEMENTATION/package_map.md — §"The src/myco/ layout" body rewritten for v0.6.0+ unified boundary (drops separate surface/, install/, mcp/, symbionts/, genesis/, meta/, providers/ blocks; adds unified boundary/ block with 4 sub-packages). Mapping matrix refreshed; "Excreted packages (historical)" section added documenting the 6-package consolidation.
• docs/architecture/L3_IMPLEMENTATION/command_manifest.md:23,29 — manifest path corrected.

Group F — Surface anchor refresh

• docs/architecture/README.md — front-page rewrite: v0.5.7 anchor → v0.6.15; layer table updated to 7 subsystems; reading order extended; governing-craft list refreshed.
• MYCO.md:64,73,86 — three v0.6.12 anchors refreshed to v0.6.15; "76 non-critical findings" preserved (count is still 76 = 9 HIGH AD1 + 67 LOW DC2/DC3/DC4/SE2).
• CONTRIBUTING.md:106 — manifest path corrected.
• docs/architecture/L1_CONTRACT/exit_codes.md:101 — v0.5.7 anchor → v0.6.15.

Group G — Scripts hygiene

• scripts/coverage_floors.py — myco/surface/ floor renamed to myco/boundary/surface/; myco/symbionts/ floor (60%) DELETED (path doesn't exist post-v0.6.0; floor never matched → dead rule).
• scripts/migrate_ascc_substrate.py:211 — manifest path corrected.

Break from v0.6.15

No backward-compat break. This is purely a hygiene molt — every change is either a path correction (where the old path was already broken) or a doctrine-doc refresh (where the old anchor was simply stale). The risk_classifier dead-pattern fix is a behavior correction (the dead patterns never fired), not an API change.

The autopoietic-loop IOU items (helper / senesce reaper persistence / auto_merge.yml) deferred from v0.6.15 craft Round 2 R-T5 / R-T6 / R-T18 are NOT closed in v0.6.16 — they are formally re-scoped into v0.6.17 (helper) and v0.6.18 (auto_merge.yml after Branch Protection setup). The split is documented in the v0.6.16 craft Round 2 + Round 3 decision.

What did NOT change

• All 7 R-rules (R1-R7): identical text.
• All 7 subsystems (Germination, Ingestion, Digestion, Circulation, Homeostasis, Cycle, Boundary): identical doctrine.
• All 46 lint dimensions: identical roster, identical severities, identical fixability.
• All 20 verbs: identical manifest, identical CLI, identical MCP shape.
• system.llm_policy: forbidden default: unchanged.
• system.governance.auto_evolve_force_high_risk: false (v0.6.15 setting): unchanged.
• system.governance.auto_evolve_pr_window_skip: false (v0.6.15 setting): unchanged.
• Schema v2 shape: unchanged.

Governing craft: docs/primordia/v0_6_16_neat_freak_sweep_craft_2026-04-29.md.
Predecessor: v0.6.15 (Agent-First default for Cycle 自起 闭环), shipped 2026-04-29.