planforge-step4-completeness-sweep

name	planforge-step4-completeness-sweep
description	Pipeline Step 4 — Completeness sweep to eliminate TODOs, mocks, stubs, and placeholder code before review.
metadata	{"author":"plan-forge","source":".github/prompts/step4-completeness-sweep.prompt.md"}

---

description: "Pipeline Step 4 — Completeness sweep to eliminate TODOs, mocks, stubs, and placeholder code before review."

Step 4: Completeness Sweep

Pipeline: Step 4 of 5 (Session 2 — Execution)
When: After all slices pass (Step 3), before the Review Gate
Model suggestion: Any model / Copilot Auto (10% token savings) — pattern scanning works well on all models
Next Step: step5-review-gate.prompt.md (new session)

Replace <YOUR-HARDENED-PLAN> with your hardened plan filename.

---

Read these files first:

1. docs/plans/AI-Plan-Hardening-Runbook.md (Section 6.1)
2. docs/plans/.md (Definition of Done)
3. .github/copilot-instructions.md

Now act as a COMPLETENESS SWEEP AGENT (see Section 6.1 of the runbook).

Scan ALL files created or modified during this phase for:

• TODO, HACK, FIXME comments
• Mock/placeholder/stub data (hardcoded records, fake values)
• "will be replaced" / "Simulate" / "Seed with sample" comments
• Stub implementations (methods that return defaults / do nothing)
• Commented-out code with future intent

For each finding:

1. Wire it to the real service/API/method
2. Remove the deferred-work comment
3. Verify build + tests pass after each batch

Output:

1. Findings count (before → after)
2. Files modified
3. New methods/types added
4. Build: pass/fail
5. Tests: pass/fail

If ANY finding cannot be resolved without scope expansion: pause and report the blocker.

---

MCP Tools (if Plan Forge MCP server is running)

• Run sweep: call forge_sweep for structured TODO/FIXME/stub scanning with file locations
• Verify scope: call forge_diff with the plan file to confirm all changes are within scope before handoff to review

Use MCP tools for structured results when available. Fall back to manual grep if MCP is not configured.

---

Persistent Memory (if OpenBrain is configured)

• Before sweeping: search_thoughts("common stubs", project: "TimeTracker", created_by: "copilot-vscode", type: "bug") — check if prior sweeps found recurring stub patterns so you can look for them proactively
• After sweep: capture_thoughts([...findings], project: "TimeTracker", created_by: "copilot-vscode", source: "plan-forge-step-4-sweep", type: "convention") — batch capture patterns and conventions discovered during cleanup