Nw BugfixSAFE

AI agents that guide you from idea to working code, with you in control at every step.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: nw-bugfix
description: "Bug fix workflow: root cause analysis → user review → regression test + fix via TDD"
user-invocable: true
argument-hint: '[bug-description] - Describe the defect observed'
---

# NW-BUGFIX: Defect Resolution Workflow

**Wave**: CROSS_WAVE
**Agents**: Rex (nw-troubleshooter) → selected crafter (OOP or FP per project paradigm)

## Overview

End-to-end bug fix pipeline: diagnose root cause, review findings with user, then deliver regression tests that fail with the bug and pass with the fix. Ensures every defect produces a test that prevents recurrence.

## Flow

```
INPUT: "{bug-description}"
  │
  ├─ Phase 1: Root Cause Analysis (@nw-troubleshooter)
  │   └─ /nw-root-why "{bug-description}"
  │   └─ Output: RCA document with root cause chain + fix proposal
  │
  ├─ Phase 2: User Review (INTERACTIVE — STOP here)
  │   └─ Present RCA findings to user
  │   └─ User confirms root cause + approves fix direction
  │   └─ If user rejects → refine RCA or stop
  │
  └─ Phase 3: Regression Test + Fix
      └─ /nw-deliver "fix-{bug-id}" — roadmap-based bugfix flow
      └─ Paradigm detection determines crafter (OOP or FP)
      └─ Regression test (RED) → fix (GREEN) → verify (COMMIT)
```

## Execution Steps

### Phase 1: Root Cause Analysis

**Skill loading**: The troubleshooter loads its skills from `~/.claude/skills/nw-{skill}/SKILL.md`:
- `nw-five-whys-methodology` — core investigation methodology
- `nw-investigation-techniques` — systematic debugging patterns
- `nw-post-mortem-framework` — structured incident analysis

Invoke @nw-troubleshooter via Agent tool:

```
Execute *investigate-root-cause for the following defect:

{bug-description}

Configuration:
- investigation_depth: 5
- multi_causal: true
- evidence_required: true

Produce:
1. Root cause chain (5 Whys with evidence at each level)
2. Contributing factors
3. Proposed fix with specific code changes
4. Files affected
5. Risk assessment of the fix
```

After the troubleshooter returns, present findings to the user. Include:
- Root cause summary (1-2 sentences)
- Evidence chain
- Proposed fix
- Files to modify
- Risk level

**STOP and wait for user confirmation before proceeding to Phase 3.**

### Phase 2: User Review

Present the RCA findings and ask:
1. "Does this root cause match your understanding?"
2. "Do you approve the proposed fix direction?"
3. "Any additional constraints or context?"

If user rejects:
- Refine the RCA with additional context
- Or stop the workflow entirely

If user approves → proceed to Phase 3.

### Phase 3: Regression Test + Fix

Phase 3 dispatches the fix through the roadmap-based bugfix flow. It does
paradigm detection (reads project CLAUDE.md for `## Development Paradigm`),
crafter selection (@nw-software-crafter for OOP, @nw-functional-software-crafter
for FP), DES enforcement, and reads the rigor profile from
`.nwave/des-config.json`.

**Preparation:**

1. Derive feature-id: `fix-{kebab-case-bug-summary}` (max 5 words)
2. Create `docs/feature/{feature-id}/deliver/` directory
3. Prepare RCA context from Phase 1 output (root cause, files affected, proposed fix)

Delegate to `/nw-deliver`:

```
/nw-deliver "fix-{bug-summary}"
```

The deliver orchestrator builds a minimal two-step roadmap:

**Step 01-01: Regression test (RED)**
- Write a test that reproduces the exact defect
- Test MUST fail against current code (proves the bug exists)
- Test location: `tests/regression/{component}/` or `tests/bugs/`
- Test name: `test_bug_{description}.py`

**Step 01-02: Fix implementation (GREEN)**
- Implement the minimal fix identified in RCA
- Run ALL tests — regression test must now PASS
- Existing tests must not regress

The crafter handles the TDD cycle (3-phase canon RED → GREEN → COMMIT per
ADR-025) with DES monitoring.

## Success Criteria

- [ ] Root cause identified with evidence at each causal level
- [ ] User reviewed and approved fix direction
- [ ] Regression test written that fails with the bug
- [ ] Fix implemented that makes the regression test pass
- [ ] All existing tests still pass (no regressions)
- [ ] Commit with conventional message: `fix(scope): description`

## Examples

### Example 1: Runtime crash
```
/nw-bugfix "DES hook crashes with FileNotFoundError when template schema is missing"
```
Phase 1: Rex traces to missing `step-tdd-cycle-schema.json` in plugin cache.
Phase 2: User confirms.
Phase 3: `/nw-deliver "fix-missing-template-schema"` → crafter writes `test_bug_missing_template_schema.py` (RED), adds fallback path resolution (GREEN), commits.

### Example 2: Silent failure
```
/nw-bugfix "Skills plugin reports success but installs zero files when source has nw-prefixed layout"
```
Phase 1: Rex traces to `is_public_skill()` returning False for all nw-prefixed names due to ownership map key mismatch.
Phase 2: User confirms.
Phase 3: `/nw-deliver "fix-ownership-map-keys"` → crafter writes regression test with nw-prefixed fixture (RED), fixes ownership map keys (GREEN), commits.

### Example 3: Functional project bug
```
/nw-bugfix "Pipeline composition breaks when filter predicate returns None"
```
Phase 1: Rex traces to missing None guard in compose() function.
Phase 2: User confirms.
Phase 3: `/nw-deliver "fix-compose-none-guard"` → paradigm detected as FP → @nw-functional-software-crafter writes property-based test (RED), adds None guard (GREEN), commits.

## Notes

- This command is for **known defects** (something is broken). For new features, use `/nw-deliver`.
- The regression test is the primary deliverable — it prevents the bug from recurring.
- Keep the fix minimal. Refactoring belongs in `/nw-refactor`, not here.
- If the RCA reveals a design flaw (not just a code bug), escalate to `/nw-design` before fixing.
- Phase 3 delegates to `/nw-deliver`, which handles paradigm detection, DES enforcement, and rigor profile automatically.

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.