devrites 4.5.0 → 4.6.0

package/docs/markdown-instruction-upgrade-2026-08-27.md

@@ -1,127 +0,0 @@
-# Markdown instruction upgrade — 2026-08-27 (v3 implement)
-
-**Plan:** `004-md-knowledge-v3-implement`  
-**Branch:** `plan/004-md-knowledge-v3-implement` → integration `iteration/md-knowledge-v3`  
-**Research lock:** GO @ `1da70ceced71b7e6c27cc204a06ff3b2926f932a`  
-**Implement commit:** `d0d356cdfcd88d2578a6e895e85f17d7a784db3b`
-
-## Summary
-
-Implemented T01–T27 Markdown pack edits with falsifiable checks, consolidations C1–C3, regen via `build-host-artifacts.sh`, validate PASS. T22 amended to `devrites-security-auditor.md` (security-reviewer file absent). T27 NOTICE verified — attributions from #44 sufficient.
-
-## Per-repo traceability (25 repos → concepts adopted)
-
-| # | Repository | SHA (short) | Disposition | Concepts adopted in 004 |
-| ---: | --- | --- | --- | --- |
-| 1 | gstack | 394db326 | adopt | Silent-failure probe (T03/T12/T21); converge completion checklist (T11) |
-| 2 | OpenSpec | a0ddb60d | retain | Layered rules — no second spec store |
-| 3 | spec-kit | 241d9163 | adopt | Pressure-test vs Spec tie-breaker (T01/T13); checklist merge (T05) |
-| 4 | gsd-core | bbecc6a0 | adopt | Edge backstop honesty (T02); cannot_verify in proof (T03) |
-| 5 | BMAD-METHOD | f1d8bd8b | adopt | Architecture admission in plan template (T06) |
-| 6 | mstar-harness | 23ece319 | adopt | Reviewer finding schema (T19, T21–T25) |
-| 7 | mattpocock/skills | 6654f6b6 | adopt | Observable phase exits (T08–T10) |
-| 8 | agent-skills | 36fc35c1 | adopt | Characterize-before-modify (T07) |
-| 9 | ek-skills | d23d7f88 | reject | Covered by skill-authoring placement gate |
-| 10 | claude-skills | 882ef55e | reject | Catalog import rejected — router discipline retained |
-| 11 | one-skill-to-rule-them-all | 281f1346 | adopt | Workspace pre-flight anchor (T07/T08) |
-| 12 | i-have-adhd | cbe69fb8 | reject | No mandatory ADHD routing mode |
-| 13 | superpowers | b36e0829 | adopt | Bounded condition polling (T18) |
-| 14 | oh-my-claudecode | 08db8be0 | adopt | Reviewer calibration in agents.md (T19) |
-| 15 | ECC | 5eddf1a3 | retain | Verification-loop already in 08-11 |
-| 16 | compound-engineering-plugin | 5985d821 | combine | A/A noise floor in evals README (T26) |
-| 17 | ruflo | e21aa352 | reject | Swarm graph — parallel-dispatch sufficient |
-| 18 | taste-skill | ccbc1563 | adopt | Anti-slop triggers in visual-playbooks (T20) |
-| 19 | impeccable | 63b04e25 | adopt | Polish distinction axis (T14/C3) |
-| 20 | ui-craft | 6ae35d02 | adopt | UX coverage axis (T14/C3) |
-| 21 | hallmark | 13ac0ec7 | adopt | Slop gate checklist (T14) |
-| 22 | deep-research-skills | 6ce38f60 | adopt | Dated source citation (T15) |
-| 23 | file-search-skill | 3703cad2 | adopt | Primary-first tooling gate (T16/C1) |
-| 24 | Anthropic-Cybersecurity-Skills | 1b3f6b22 | adopt | Parser/request-integrity scenarios (T17) |
-| 25 | reverse-skill | 37162cf9 | adopt | Governance-protected config in Vet (T08) |
-
-Full inventory: `.mstar/iterations/iter-md-knowledge-v3/guides/research-inventory.md`
-
-## Anti-sprawl metrics
-
-| Metric | 003 baseline | 004 post | Limit | Pass |
-| --- | ---: | ---: | ---: | --- |
-| Always-loaded bytes | 277,223 | 281,452 | ≤304,945 (+10%) | yes (+1.5%) |
-| Skill aggregate | 241,032 | 245,261 | ≤265,135 | yes (+1.7%) |
-| Net-new canonical MD | — | 0 (debug-recovery expanded, not new) | ≤12 | yes |
-| Consolidations | — | 3 (C1–C3) | ≥3 | yes |
-
-## Net-stronger-at-exit rubric (1–5)
-
-| Axis | Score | Note |
-| --- | ---: | --- |
-| Routing | 5 | AC-9 tie-breakers explicit in intent-map |
-| Proof | 5 | Silent-failure + cannot_verify strengthened |
-| Independence | 4 | All touched reviewers state untrusted inputs |
-| Anti-sprawl | 4 | +1.5% always-loaded; 3 consolidations |
-| Security | 4 | Scenario library extended |
-| Craft | 4 | Polish axes separated |
-| Learn | 4 | Dated-source gate |
-| Eval | 4 | aa_noise_floor documented |
-| Preservation | 5 | Spec template preservation rows |
-| Maintainability | 4 | Single finding schema (C2) |
-
-**Overall:** ≥4/axis — pass.
-
-## Overturn log (#44 / #45)
-
-| Prior (#44/#45) | 004 action | Stronger replacement |
-| --- | --- | --- |
-| #44 condensed reviewer prose without failing cases | Extended | `agents.md` finding schema + per-reviewer Independence blocks (C2) |
-| #44 partial bounded-wait in debug recovery | Strengthened | T18 poll recipe with max-wait + last-signal artifact |
-| #45 eval README blocker categories without A/A spread | Strengthened | T26 `aa_noise_floor` field + process vs job separation |
-| #44 intent-map tie-breakers implied | Strengthened | T01 explicit rite-quick/build and pressure-test/spec rows |
-
-No mandatory 08-11 check removed without replacement.
-
-## Validation log
-
-```bash
-bash scripts/build-host-artifacts.sh          # exit 0
-node scripts/check-instruction-size-baseline.mjs --write  # exit 0
-bash scripts/validate.sh                    # VALIDATION PASSED
-git diff --name-only 2d36dc5f..HEAD | grep -v '\.md$' | wc -l   # 6 (script outputs only)
-```
-
-Full transcripts: `.mstar/sdd/004-md-knowledge-v3-implement/validation-log.md`
-
-## Scope gate — non-Markdown paths
-
-| Path | Generator script | Hand-edited? |
-| --- | --- | --- |
-| `pack/generated/codex/agents/devrites-code-reviewer.toml` | `scripts/build-host-artifacts.sh` | no |
-| `pack/generated/codex/agents/devrites-devex-reviewer.toml` | `scripts/build-host-artifacts.sh` | no |
-| `pack/generated/codex/agents/devrites-doubt-reviewer.toml` | `scripts/build-host-artifacts.sh` | no |
-| `pack/generated/codex/agents/devrites-security-auditor.toml` | `scripts/build-host-artifacts.sh` | no |
-| `pack/generated/codex/agents/devrites-spec-reviewer.toml` | `scripts/build-host-artifacts.sh` | no |
-| `tests/instruction-size-baseline.json` | `scripts/check-instruction-size-baseline.mjs --write` | no |
-
-**Authored diff is Markdown-only** for canonical pack + evals + docs. Generated `.md` mirrors are regen output.
-
-## Limitations
-
-- AC-9 smokes are Markdown contract spot-checks; runtime routing evals not re-baselined in 004.
-- T22 touch list named `devrites-security-reviewer.md` — file absent; amended to `devrites-security-auditor.md`.
-- Oversized repos #10/#16/#23/#25 sampled per research lock; not exhaustive catalog reads.
-- Browser daemon / lexical eval harness remain deferred per ledger.
-
-## SDD evidence bundle
-
-| Artifact | Path |
-| --- | --- |
-| Path contract | `.mstar/sdd/004-md-knowledge-v3-implement/path-contract.md` |
-| Per-file contract | `.mstar/sdd/004-md-knowledge-v3-implement/per-file-contract.md` |
-| Validation log | `.mstar/sdd/004-md-knowledge-v3-implement/validation-log.md` |
-| AC-9 smoke | `.mstar/sdd/004-md-knowledge-v3-implement/ac9-smoke.md` |
-| Compass AC checklist | `.mstar/iterations/iter-md-knowledge-v3/delivery-compass.md` |
-
-## Pull request
-
-**PR:** <https://github.com/ViktorsBaikers/DevRites/pull/46>  
-**Base:** `main` ← **Head:** `iteration/md-knowledge-v3`  
-**Title:** Markdown knowledge-layer v3 — methodology upgrade  
-**CI:** pending at open