carmandale

rp-cli-investigate

"Deep investigation workflow using rp-cli and RepoPrompt's context builder. Use when investigating bugs, understanding codebases, or tracing issues through complex systems. CRITICAL: Do not timeout rp-cli commands."

carmandale 2 Updated 3mo ago
GitHub

Install

npx skillscat add carmandale/agent-config/rp-cli-investigate

Install via the SkillsCat registry.

SKILL.md

Deep Investigation Mode (rp-cli)

Systematic investigation workflow using RepoPrompt's context building capabilities.

⚠️ CRITICAL: NO TIMEOUTS

Do NOT add timeouts to rp-cli commands.

The builder command invokes AI-powered context analysis that takes 60-90+ seconds. This is normal. Let it complete.

Wrong:

# ❌ Timeout → "didn't work" → skip to manual file reading
timeout 30 rp-cli -e 'builder "..."'

Right:

# ✅ Wait for actual completion
rp-cli -e 'builder "Investigate the auth flow"'

If you bail on builder because it "took too long", you are doing it wrong.


Quick Reference

# Core exploration
rp-cli -e 'tree'                              # File tree
rp-cli -e 'search "pattern"'                  # Search code
rp-cli -e 'read path/file.swift'              # Read file
rp-cli -e 'structure path/'                   # Code structure

# Context building (THE MAIN TOOL - wait for it!)
rp-cli -e 'builder "instructions" --response-type plan'

# Follow-up in same tab
rp-cli -t '<tab_id>' -e 'chat "follow-up question" --mode plan'

Investigation Protocol

Phase 1: Initial Assessment (BRIEF)

  1. Read any provided files/reports (traces, logs, error reports)
  2. Summarize symptoms and constraints
  3. Form initial hypotheses

Do NOT do extensive file reading here. Save deep exploration for builder.

Phase 2: Systematic Exploration via builder (REQUIRED)

⚠️ You MUST call builder. Do not skip this step.

rp-cli -e 'builder "Investigate: <specific area>

Symptoms observed:
- <symptom 1>
- <symptom 2>

Hypotheses to test:
- <theory 1>
- <theory 2>

Areas to explore:
- <files/patterns/subsystems>
" --response-type plan'

Wait for this to complete. It will take 60-90+ seconds. That's normal.

Phase 3: Follow-up Deep Dives

After builder returns, continue with targeted questions:

rp-cli -t '<tab_id>' -e 'chat "<specific follow-up>" --mode plan'

Phase 4: Evidence Gathering

  • Check git history for recent relevant changes
  • Look for patterns across similar files
  • Trace data/control flow through the codebase

Phase 5: Conclusions

Document:

  • Root cause identification (with evidence)
  • Eliminated hypotheses (and why)
  • Recommended fixes

Anti-Patterns (NEVER DO)

Anti-Pattern Why It's Wrong
Adding timeouts to rp-cli Commands legitimately take time; timeout ≠ failure
Skipping builder You'll miss critical context the AI analysis provides
5+ manual file reads before builder Save deep reading for after context is built
Declaring "builder didn't work" after timeout It was working; you killed it
Falling back to manual investigation The whole point is to use builder for context

Report Template

# Investigation: [Title]

## Summary
[1-2 sentence summary]

## Symptoms
- [Observed symptom 1]
- [Observed symptom 2]

## Investigation Log

### Phase - [Area Investigated]
**Hypothesis:** [What you were testing]
**Findings:** [What you found]
**Evidence:** [File:line references]
**Conclusion:** [Confirmed/Eliminated/Needs more investigation]

## Root Cause
[Detailed explanation with evidence]

## Recommendations
1. [Fix 1]
2. [Fix 2]