v5.16.2 - ClawScan follow-up

ClawHub's LLM review of 5.16.1 found three issues in the bundled Python scripts. All are fixed:

ClawScan follow-up (ClawHub skill)

ClawHub's LLM review of 5.16.1 found three issues in the bundled Python scripts. All are fixed:

Security

• Environment isolation (--env): blackboard.py --env moved the blackboard but kept pending changes and the lock in the startup environment, so a development proposal could be committed to the production blackboard. One env-scoped directory now drives the blackboard, pending changes, and lock. check_permission.py --env now also moves the signing key.
• Saved-context prompt safety: context_manager.py skipped validation of updated_at yet printed it inside the injected <project_context> block. It is now validated as an ISO-8601 timestamp and printed only when it is one (even with --force).
• Audit summary token masking: --audit-summary --json returned full grant tokens and HMAC signatures. Tokens are now masked to a 16-character prefix and signatures omitted.

Regression checks reproduce each finding against 5.16.1 and pass on 5.16.2. Full suite: 3,888 tests across 42 suites; tsc --noEmit clean.

See CHANGELOG.