Use when asked to verify, improve, or sharpen a plan file using deep plan verification. Runs the deep-plan-verification babysitter process (yolo mode) on a plan file — no coding, no implementation, only iterative plan QA. Shows before/after diff and final quality score at the end.
Announce at start: "I'm using the deep-verify-plan skill to run iterative plan verification."
Runs the deep-plan-verification.js babysitter process against a plan file. This does zero implementation — it only improves the plan document itself through iterative QA:
/deep-verify-plan [path-to-plan.md]
If no path is given, auto-detect the most recently modified plan in docs/superpowers/plans/.
Step 1: Find the plan file
If user provided a path, use it. Otherwise:
ls -t docs/superpowers/plans/*.md 2>/dev/null | head -1
Confirm with the user: "Found plan: <path>. Running deep verification on it."
If no plan found anywhere, ask the user to provide the path.
Step 2: Snapshot the plan (before state)
cp "<plan-file>" "<plan-file>.before-deep-verify-plan"
Step 3: Invoke babysitter yolo
Use skill babysitter:yolo:
/babysitter:yolo Run the deep-plan-verification process.
Process file: ~/.a5c/processes/deep-plan-verification.js
Inputs: planFile=<absolute-path-to-plan>, projectRoot=<CWD>, requireApproval=false, taskDescription=Deep plan verification — improve plan quality to 95/100 across 8 dimensions
CRITICAL: Do NOT write any code, create any files beyond the plan update, or implement anything. This run is plan-only.
Step 4: Show the diff (after state)
After babysitter completes:
diff "<plan-file>.before-deep-verify-plan" "<plan-file>" || true
If no diff: "No changes were made to the plan."
If diff exists, present it clearly:
Step 5: Clean up snapshot
rm "<plan-file>.before-deep-verify-plan"
Step 6: Report to user
Summarize:
.md file itself