| name | flywheel-reality-check |
| description | Strategic gap analysis between vision (AGENTS.md / README.md / plan docs) and what's actually implemented. Converts gaps into beads, optionally launches a swarm. Use when "reality check", "where are we really", "gap analysis", "did we drift", or before declaring a long-running project done. |
Flywheel Reality Check — direct entry point
This skill is a thin pointer to the canonical reality-check sub-flow at skills/start/_reality_check.md. It exists so users can invoke the gap-analysis pass directly via /agent-flywheel:flywheel-reality-check without going through /agent-flywheel:start → "Reality check".
When to use
- A long-running multi-agent project has accumulated dozens of closed beads — it's time to step back and verify the aggregate delivers on the original vision.
- Reviews are converging but you suspect drift between aspirational docs and actual implementation.
- Before a release / before declaring done — the "come-to-Jesus" alignment pass.
- After
/agent-flywheel:flywheel-drift-check flagged significant drift and you want the deeper strategic lens.
Instructions
-
Read skills/start/_reality_check.md end-to-end (resolve via $CLAUDE_PLUGIN_ROOT/skills/start/_reality_check.md or fall back to the cache path under ~/.claude/plugins/cache/agent-flywheel/agent-flywheel/<VERSION>/skills/start/_reality_check.md).
-
Execute its depth-selection AskUserQuestion (Reality check only / Reality check + beads / Full pipeline) and route to the matching section verbatim. Do NOT pick a depth unilaterally — per the start-skill's UNIVERSAL RULE 1, this is a labeled-option decision.
-
The slash-named skill /reality-check-for-project is load-bearing — invoke via the Skill tool when reaching Phase 1, do NOT paraphrase its prompts.
-
CASS capture (§2) and bead tagging (§3) are mandatory — they make the reality-check round queryable across future sessions.
-
If you're in the middle of an active flywheel session (checkpoint exists), the reality-check findings should produce beads tagged reality-check-<YYYY-MM-DD> that the existing swarm can pick up via flywheel_advance_wave. If standalone, the "Full pipeline" depth launches a fresh 3 cod + 3 cc swarm specifically for gap-closure (pi fallback if Codex unavailable; see AGENTS.md NTM pane priority).
Why this is a thin pointer
Same reasoning as commands/start.md (v3.6.3): single source of truth at skills/start/_reality_check.md eliminates drift between the slash-command surface and the canonical workflow. Every menu/routing change happens once, in one file.