Command
Groundtruth: whole-repo debt scan (stubs, no-ops, phantom refs, secrets, RLS-off) — findings, not a verdict.
Groundtruth — a Claude Code plugin that audits whether an agent actually did what it was asked; catches the false 'Done' on Stop (missed subtasks, stubs, false 'tests pass', overridden rules).
Command
Groundtruth: whole-repo debt scan (stubs, no-ops, phantom refs, secrets, RLS-off) — findings, not a verdict.
Command
Turn Groundtruth BLOCK mode on/off (default: warn). No settings.json edit needed.
Command
What Groundtruth checks, and its commands.
Command
Opt-in (NOT default): fan out an agent to read your rule docs and PROPOSE richer Groundtruth rules beyond the literal X not Y forms. Routed through the same grounding + /groundtruth-rules approval gate — nothing arms, and the model runs only when you invoke this.
Command
Review the rules Groundtruth extracted from your docs and approve which to arm (the permission gate).
Command
One-shot Groundtruth setup check — detects what's configured and hands you the exact actions for the rest.
Command
Show Groundtruth's latest verdict card for this session.