| name | debug+ |
| description | Invoke when debugging, fixing bugs, diagnosing failures, or when something is broken, unexpected, or not working as expected. Trigger phrases: 'debug', 'fix bug', 'why is X failing', 'what's wrong with', 'broken', 'diagnose', 'investigate'. Wraps superpowers:systematic-debugging with mandatory scout context harvesting. |
| argument-hint | [bug description, error output, or what is broken] |
| user-invocable | true |
debug+ — Systematic Debugging with Scout Context
Wraps superpowers:systematic-debugging. Requires superpowers to be installed.
No phase prerequisite — invoke from any skill chain phase when you encounter a
bug, test failure, or unexpected behavior.
Before You Begin
This skill adds scout-mandated context harvesting on top of the base debugging skill:
- Scout context — always map the affected code area via the scout agent before debugging
- Environment awareness — detects test infrastructure, dependencies, and recent changes
- Constitutional awareness — loads active enforcement rules from session context
Procedure
Phase A: Harvest Context
-
Reproduce the issue — run the failing command or test and capture the exact output.
-
Invoke the scout agent to map the affected area:
Agent(subagent_type="scout", prompt="Map the codebase around [affected files/modules]. Focus on: related functions, callers, test files, and recent changes.")
-
Identify:
- The exact failure symptom (error message, wrong output, crash)
- The code path that produces the symptom
- Recent changes that may have introduced the issue
- Related test coverage
- Active enforcement rules from session context (see session-start output)
Phase B: Investigate (delegate to superpowers:systematic-debugging)
-
Invoke superpowers:systematic-debugging with the gathered context.
-
Follow the systematic debugging process:
- State the hypothesis before investigating
- Verify the hypothesis with evidence (command output, file reads)
- If the hypothesis is wrong, state why and form a new one
- Show the evidence at each step
Phase C: Report Findings
-
Produce a finding summary:
## Investigation Report
### Symptom
[Exact failure description]
### Root Cause
[What caused the issue]
### Evidence
- [Evidence 1: command output or file content]
- [Evidence 2: ...]
### Fix
[Specific change needed to resolve the issue]
-
If the fix is straightforward, implement it and run the relevant tests.
If the fix requires a larger change, return to the current phase with the finding.
Skill Chain
/debug+ has no phase prerequisite — invoke it from any phase when you need to
debug a bug or investigate unexpected behavior. After resolving the
investigation, return to your current phase.
Completion
Report one of these states when the skill finishes:
- DONE — Root cause identified and fix applied with passing tests.
- DONE_WITH_CONCERNS — Root cause identified but fix has side effects or needs review.
- BLOCKED — Cannot reproduce the issue or insufficient context to diagnose.
- NEEDS_CONTEXT — Need user input to reproduce the issue or understand expected behavior.