Answers the 3 adoption questions (Plan-21 follow-up):
1) LEVEL SELECTION (4 packaging levels, packaging/levels.json):
- install.sh --level core|devkit; platform refused (preview service),
enterprise refused (future). Level recorded in .casan-level.
- casan init --level 1..4: L1=gate+Plan-20 hooks only; L2=+CI+domain-pack;
L3=L2 base+preview note; L4=refused. New `casan level show|set`.
- levels.json core now includes adapters/ + schemas/ + install scripts.
2) EXISTING SHELLS (agents/skills): init MERGES Plan-20 hooks into an existing
.claude/settings.json and .codex/{hooks.json,config.toml} idempotently
instead of clobbering — preserves the project's own hooks/agents/skills and
unrelated keys. Re-running never duplicates the CASAN hook.
3) NO RE-INDEX / NO SHELL REWRITE: init only adds config; it does not parse or
index code and does not rewrite the project shell.
Safety fixes after a test accidentally ran init in the real repo:
- launcher shim now SELF-LOCATES its install from its own path (no ambient
CASAN_HOME cross-talk).
- casan init REFUSES to adopt a CASAN source hub into itself (--force to
override), so the Plan-20 hooks can't block the developing agent.
- test always runs init inside throwaway dirs; +source-hub guard test.
hybrid-install-tests.sh: 41/41 PASS.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adopt CASAN like a normal tool (codegraph-style): install the harness ONCE
per machine, then `casan init` per project writes CONFIG ONLY — the harness
is no longer copied into every repo.
- install.sh / install.ps1: global bootstrap (curl|sh / irm|iex or local
source). Installs harness to $CASAN_HOME/versions/<ver>, writes a `casan`
launcher that resolves the shared harness + the current project's .specify,
and records a gate-code integrity hash. CASAN_NO_PATH_LINK for tests.
- harness_hash.py: deterministic content hash over gate code (scripts/bash,
scripts/python, security, level5) — the pin+verify anchor.
- casan-init.py: `casan init` writes .casan/{config,version.lock,agentic.env},
.specify/ marker, and the Plan-20 client hooks — no harness copy. `verify`
recomputes the harness hash LIVE and compares to the project pin (drift/
tamper -> rc 3), preserving the Plan-16 trusted-gates guarantee off-repo.
- bin/casan: new `init` and `verify-harness` commands.
- hybrid-install-tests.sh: 21/21 (install, config-only init, no-copy, pin,
verify ok, tamper drift, bridge runs against project state via global harness).
- docs: CASAN_INSTALL_HYBRID.md + Plan-21.
The path model (casan-paths.sh) already separated harness/state/domain roots,
so this is installer + init, not a core rewrite. Remote dist tarball, real
Windows run, and signed .harness-hash are the documented next steps.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wave 0 + Wave 1 core of the transparent agentic-client integration: a
developer types prompts normally in Claude Code / Codex while every
certified turn still carries a full H1->H7 trace and an H6 record.
- agentic_bridge.py: stdlib-only lifecycle state machine (begin/pre-tool/
post-tool/telemetry/finalize/abort + report/doctor). Single-model
invariant (never calls a model), fail-closed at the side-effect point,
admission TTL + canonical-project/session binding, atomic state under
.specify/state/agentic-sessions/, secret redaction, null-not-zero H6.
- agentic-lifecycle.schema.json: client-agnostic JSON contract.
- adapters/claude-code + adapters/codex: thin hook renderers + config
templates that call the core bridge.
- phase-agentic-bridge-tests.sh: C1-C12 acceptance + threat suite (30/30).
- devkit templates/{claude,codex} + windows/install-agentic.ps1
(install/doctor/uninstall with manifest, path-safe).
- docs/casan Windows + security/bypass guides; plan status -> IMPLEMENTED.
- harden generate-agentops-dashboard.py aggregation against null H6 costs.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Organize CASAN as a reusable source hub with SPLIT releases so downstream adopts only the
level it needs (packaging/levels.json is the single source of truth).
Implemented now:
- Level 1 Core: bin/casan CLI (run/gate/test/verify/reuse/dashboard) + VERSION.
- Level 2 DevKit: packages/casan-devkit (install.sh, Dockerfile.harness, templates:
project scaffold, domain-pack, gitea-workflow).
- scripts/package-release.sh core|devkit|platform|all-in-one-demo — builds split bundles
into dist/ (BUNDLE-MANIFEST + SHA256SUMS); platform is stamped PREVIEW/INCOMPLETE;
enterprise (future) is REFUSED (exit 3, no fake-complete package). Bundles verified:
extract → bin/casan works, deterministic + domain suites pass, casan reuse VALID.
- docs/packaging: CASAN_PACKAGING_PLAN + ADOPTION + CI + DOMAIN_PACK + GITEA_PACKAGE + DOCKER.
Structure + docs only:
- Level 3 packages/casan-platform (dashboard exists; control-panel/viewers pending).
- Level 4 packages/casan-enterprise (RBAC/tenant/KMS/WORM/approval exist in core; governed
console pending). No Chat Console/RBAC-console/tenant-console/model-mgmt built in this task.
Harness change (enables extracted bundles to self-resolve): casan-paths.sh + the Python
project_root() walk-ups now accept a second root marker `packages/casan-harness` in addition
to `.specify`, so a freshly-unpacked core/devkit/demo bundle (no `.specify` yet) roots
correctly and creates state on first run. In an adopted repo `.specify` still matches first.
policy-bundle.yaml paths corrected to packages/casan-harness (re-signed). Full gate 64/0/3.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
thanhnv