Harden and qualify the 21 local business skills

.agents/skills/validate-sas-release/SKILL.md

@@ -18,3 +18,13 @@ Business expansion changes must also pass `npm run check:business-skills`: the s
 For explicitly requested adversarial evaluation without Lovable generation credits, use the finite local batch in `docs/ADVERSARIAL-BUSINESS-EVALUATION.md` and the existing deterministic scorer. Run only an existing cached model with recorded version/runtime fingerprints; retain real outputs and usage. Jev is advisory; primary trace review and the original approval criteria remain authoritative. Never substitute mocks, input-helper success or truncated/error outputs for a passing model run. Record results against the exact immutable registry version/hash, dry-run insertion SQL in rollback, and publish only qualifying versions through the existing approval gate.
 
 For corrected cohorts, use a new immutable skill version and evidence directory. Run `scripts/compare-business-adversarial.mjs BEFORE_DIR AFTER_DIR NEW_REPORT_PATH` after real reports exist; changed inputs, expectations, model settings, thresholds or incomplete cohorts must fail comparison. Record actual improvement and remaining failures; never equate improvement with approval.
+
+When the owner explicitly authorizes a model migration, retain failed histories and qualify the new exact model plus packaged runtime separately. The same-model comparison must reject changed models/protocols. Preserve case inputs, expectations, thresholds and raw outputs. Follow the packaged qualification section in `docs/ADVERSARIAL-BUSINESS-EVALUATION.md`: hash-bound profiles, real routing and generation, no output repair, archive allowlist, and substantive valid-domain review. Never rewrite a registered version or its runtime while a run is active.
+
+Publication plans must reject empty/no-op targets and stale source/archive/run bindings. Native SQL is not an authenticated administrator session: use the existing admin review flow for publication, preserving `enforce_admin_only_publish` and the adversarial trigger. Do not forge claims or alter roles/triggers to publish. Verify actual public versions and model/protocol scope afterward.
+
+The original business case contract is pinned in `scripts/adversarial-approval.mjs`. Qualification, registry recording and publication planning require all 21 skills and 273 unchanged cases, including inputs, contexts, severities and expectations. Pilots never qualify. Test changes are not an allowed failure fix. Review each distinct raw response against the original domain expectation and record hash-bound decisions; group identical baseline evidence with the existing review-group command without automatically generating verdicts.
+
+For interrupted local inference, reuse the worker's explicit `--resume` only when it accepts the original incomplete JSONL prefix and exact plan/runtime/package fingerprints. Preserve all prior rows, including errors. The worker requires a hash-bound attempt journal ending at a completed case, appends only not-yet-started cases and records its actual resumption. An in-flight or unknown attempt blocks recovery. Use `--max-cases N` for an explicit boundary checkpoint; exit 4 remains incomplete evidence. Changed or legacy runtime evidence must remain separate; never edit a raw output or its plan hash to make recovery pass. Run the deterministic recovery regressions with `python3 tests/worker_resume_test.py`.
+
+When diagnosing a partial pilot, use the existing evaluation command's `diagnose NEW_JSON --workdir DIR` mode. Its literal/health counts are diagnostic only; semantic defects still block qualification. Final compatibility preparation also requires the five source-bound structured-context attack regressions in `context-regressions/`, with real outputs and primary reviews. The preparation SQL inserts only model-scoped compatibility evidence; publication flags and the actual reviewer are set exclusively through the authenticated admin flow.

.gitignore

@@ -42,3 +42,7 @@ artifacts/validation/
 
 # Private raw adversarial traces and judge evidence
 artifacts/adversarial-business/
+
+# Local Python caches are not skill source artifacts.
+__pycache__/
+*.pyc

CHANGELOG.md

@@ -11,6 +11,16 @@ Component artifacts are versioned and released independently via git tags:
 - **`super-agent` CLI** — `cli-v*` tags publish cross-platform binaries.
 - **`@superagentskill/sdk`** — versioned in `packages/sdk-ts/package.json`.
 
+## [0.6.4] — 2026-10-23 — *Genesis* (alpha) — Maintenance release
+
+### Changed
+
+- Add a pinned local-model runtime with recorded routing, raw generation and exact model/SDK/archive bindings.
+- Bind business qualification to the original 273-case contract; keep pilots and failed immutable versions separate from publication.
+- Exclude local requests, inference outputs and Python caches from business package archives.
+- Preserve completed inference checkpoints with an attempt journal; block silent retries of interrupted cases.
+- Patch dependency advisories and migrate record validation to Zod 4, with an actual MCP schema/execution regression.
+
 ## [0.6.3] — 2026-10-23 — *Genesis* (alpha) — Maintenance release
 
 ### Changed