Plugin Health

Run on-demand integrity check for a plugin. Compares .install-manifest.json SHA256 entries against actual files on disk. Detects tampering, partial corruption, or missing files.

How to use

Step 1 — Resolve the slug

If not given, list and ask:

from dashboard.backend.sdk_client import evo

plugins = evo.get("/api/plugins")

Step 2 — Run health check

health = evo.get(f"/api/plugins/{slug}/health")

Response shape:

{
  "status": "active" | "broken" | "routine_activation_pending",
  "reason": "sha_mismatch" | "missing_files" | null,
  "files": [".claude/agents/plugin-foo-agent.md", ...],  // if broken
  "checked_at": "..."
}

Step 3 — Report

If active: plugin is healthy. Summarise: total files checked, all SHA256 matching, registered hooks/heartbeats/routines still wired.

If broken:

• List corrupted/missing files
• Suggest options:
  • Reinstall: plugin-install with same source URL (overwrites files, preserves data)
  • Uninstall + reinstall: clean slate
  • Manual fix: edit files back to original state (if user knows what was changed)

If routine_activation_pending: scheduler was offline during install. Tell user to start/restart scheduler.

Error handling

• HTTP 404 not_found — plugin not installed.
• HTTP 500 health_check_failed — filesystem permission issue or manifest unreadable. Check ADWs/logs/plugins/.

Notes for the agent

• Broken state doesn't auto-disable. Plugin files may still work partially. User should decide whether to act.
• If multiple plugins broken at once, likely the user did a git operation or .claude/ migration. Health check is diagnostic, not autofix.