Point the canonical entry points at the new package layout; the .specify compat facade
is retained as an intentional backward-compat layer (see below).
- .gitea/workflows/{ci,harness-ci}.yml: invoke packages/casan-harness/scripts/bash|tests
(state paths .specify/logs kept). CI now runs on the new structure.
- infra/local-prod/docker-compose.yml: dashboard-server.py -> packages path (logs/alerts
env stay under .specify state).
- scripts/casan-step.mjs + run-casan-pipeline.mjs: resolve the harness under
packages/casan-harness, falling back to .specify so the adversarial/sourcegen sandboxes
(which stage a .specify/ tree) keep working; requirement input prefers apps/okr/domain/input.
- project-registry.json: record the new layout (harness_root, state_root, governance_root,
per-project domain_root) so Plan-06 can register a second app with its own domain.
verify-harness-reuse.sh already resolves via CASAN_HARNESS_ROOT -> HARNESS_REUSE_VALID (3 projects).
Facade decision: the .specify/{scripts,tests,security,...} symlinks are KEPT as a
documented compat layer. A full hard cutover (removing them) still needs ~15 literal
`.specify/...` refs repointed (loop_common/evidence-pack/secrets-scan config+test paths,
run-casan-pipeline step scripts, and the signed policy-bundle.yaml path list which then
needs manifest regen + re-sign). That is a scoped follow-up; the physical separation
(code in packages/, domain in apps/, packages holds no domain data, single CASAN_* path
indirection) is complete and the full gate is green via BOTH entry paths: PASS=64 FAIL=0 SKIP=3.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a shared log taxonomy (error<warn<info<debug<trace, default info) wired
through all three layers, plus a machine-readable per-step run log.
- scripts/casan-log.mjs + .specify/scripts/bash/casan-log.sh: shared logger
([LEVEL] ts [component] msg), stderr-only so stdout contracts and exit codes
are byte-identical. Node side includes trace-level redaction (secret/PII).
- run-casan-pipeline.mjs: per-STEP info line, loop-activation warnings
(BACK-TO-PLAN, FAIL->STEP6), debug harness rc + trace_id, trace payload
excerpt, end-of-run 13-STEP summary table, and one JSONL line/step in
.specify/logs/pipeline-run.jsonl. New --dry-run stubs all agents but keeps
the real wrapper in the loop (deterministic, offline, full 13 STEP + loops).
- casan-step.mjs: debug logs for judge verdict, checkpoint, rollback; logger
import degrades to noop when the file is copied standalone (T1 test).
- casan-harness.sh: debug-log each phase H4-in -> H5 -> [H2-gate] -> H6-exec
-> H4-out with its rc; optional CASAN_PHASE_REPORT JSON for the Boss.
Default level (info) keeps output close to before; behavior opt-in via env.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Hướng A — scorecard.sh (video demo):
- h5_1 approval workflow: hardcode 0 → governance-check deploy live (approval_required)
- h6_2 hallucination rate: hardcode 0 → hallucination-scan phân biệt dirty>clean live
- "N/5 mục" chuyển từ text cứng sang đếm động
- H4/H5/H6 → 100/100 (5/5 gate live), Average 57.9 → 90.0
Hướng B — run-casan-pipeline.mjs:
- fallback: stub 'exit 9' → 'cat /nonexistent' (real failure, nhất quán adversarial T3)
- drift: giữ so fallback-output vs golden (clean run=1.0); năng lực phát hiện
drift thật chứng minh ở adversarial suite
- Full 12-step run verify: H1 CONTEXT_VALID=24, H2 tool-audit records=25 signed,
H5 audit-chain records=22 signed, H6 provider_telemetry per-step thật, H7 rollback real
phase3-real-run-scoring.md: giải thích vì sao scorecard cũ cho H5=60/H6=80
(hardcode), phân biệt scorecard-90 vs re-score-84 (2 mục đích khác nhau).
Verify: adversarial 44/0, security-gate 11/0/0, pipeline 12 steps OK.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
