All hardcoded 'clawdie' references in production code now derive from
AGENT_NAME (default: 'clawdie'). This makes the mevy canary strategy
reliable — changing AGENT_NAME is all that's needed.
Changes:
- Hardcoded paths: CMS_WEBROOT, ASTRO_SITE_PATH, verify checks,
controlplane dashboard dir, sessions dir, output dir, chown user
- Prometheus metrics: prefixed with AGENT_NAME for multi-install dashboards
- hostd log strings: use AGENT_NAME instead of 'clawdie-hostd'
- MCP server name: derived from AGENT_NAME
- Skill modify patches: container image and mount allowlist use AGENT_NAME
- SQL migration file renamed: clawdie-brain-hybrid-upgrade → brain-hybrid-upgrade
- Temp dir prefixes: all use AGENT_NAME
Kept as-is (correct pattern):
- 'clawdie' as default fallback when AGENT_NAME is unset
- .pi/extensions/clawdie-harness/ directory (pi package identity)
- html/docs-clawdie-si/ (public docs site URL)
---
Build: pass | Tests: pass — 1527 passed, 3 failed (2 files, pre-existing)
- Add permissions.sql with GRANT + ALTER DEFAULT PRIVILEGES
- Add fix-permissions.sh for quick repair on existing databases
- Add POSTGRES-PERMISSIONS.md documentation
- Update setup-db-jail.sh to run permissions.sql after migrations
- Update skills-memory.ts to apply permissions after artifact import
Fixes: permission denied for table memories
