From oh-my-auggie
2-stage pipeline: trace (causal investigation) -> deep-interview (requirements crystallization) with 3-point injection
npx claudepluginhub r3dlex/oh-my-auggie --plugin oh-my-auggieThis skill uses the workspace's default tool permissions.
<Purpose>
Verifies tests pass on completed feature branch, presents options to merge locally, create GitHub PR, keep as-is or discard; executes choice and cleans up worktree.
Guides root cause investigation for bugs, test failures, unexpected behavior, performance issues, and build failures before proposing fixes.
Writes implementation plans from specs for multi-step tasks, mapping files and breaking into TDD bite-sized steps before coding.
Share bugs, ideas, or general feedback.
<Use_When>
<Do_Not_Use_When>
/oma:deep-interview directly/oma:trace directly<Execution_Policy>
/oma:deep-interview protocol (one question at a time)state_write(mode="deep-interview") with source: "deep-dive"{{ARGUMENTS}}Explore agentstate_write(mode="deep-interview")Present the 3 hypotheses to the user via AskUserQuestion:
Starting deep dive. I'll investigate your problem through 3 parallel trace lanes, then use the findings for a targeted interview.
Your problem: "{initial_idea}" Proposed trace lanes:
- {hypothesis_1}
- {hypothesis_2}
- {hypothesis_3}
Are these hypotheses appropriate?
Options:
Run 3 parallel tracer lanes. Each lane gathers evidence FOR and AGAINST its hypothesis, names the critical unknown, and recommends a discriminating probe.
Trace Output Structure (saved to .oma/specs/deep-dive-trace-{slug}.md):
# Deep Dive Trace: {slug}
## Ranked Hypotheses
| Rank | Hypothesis | Confidence | Evidence Strength |
|------|------------|------------|-------------------|
| 1 | ... | High/Med/Low | Strong/Moderate/Weak |
## Per-Lane Critical Unknowns
- **Lane 1**: {critical_unknown}
- **Lane 2**: {critical_unknown}
- **Lane 3**: {critical_unknown}
## Most Likely Explanation
[current best explanation]
## Recommended Discriminating Probe
[next probe to collapse uncertainty]
Follow /oma:deep-interview protocol with 3 initialization overrides:
Override 1 — initial_idea enrichment:
Original problem: {ARGUMENTS}
<trace-context>
Trace finding: {most_likely_explanation}
</trace-context>
Given this root cause, what should we do about it?
Override 2 — codebase_context replacement: Skip deep-interview's explore step. Inject trace synthesis as codebase_context.
Override 3 — initial question queue: Extract per-lane critical_unknowns from trace result. Ask these FIRST, then continue with normal ambiguity-driven questioning.
Low-confidence handling: If all lanes low-confidence, use original user input without enrichment.
Follow /oma:deep-interview ambiguity scoring (goal 40%, constraints 30%, criteria 30%) until ambiguity ≤ 20%.
Generate spec in standard deep-interview format with additional section:
## Trace Findings
[summarizes trace results that shaped the interview]
Save to .oma/specs/deep-dive-{slug}.md.
Question: "Your spec is ready (ambiguity: {score}%). How would you like to proceed?"
Options:
IMPORTANT: Invoke via Skill(). Do NOT implement directly.
<Tool_Usage>
Explore agent for brownfield codebase explorationAskUserQuestion for lane confirmation and interview questionsstate_write / state_read for state persistenceWrite tool to save trace result and specSkill() to bridge to execution modes<trace-context> delimiters
</Tool_Usage><Final_Checklist>
<trace-context> delimitersmode="deep-interview" with source: "deep-dive" discriminator
</Final_Checklist>