Investigate

Run before starting any implementation task. Assembles all relevant context in one call.

Arguments

• <task_description> — what you're about to implement (natural language)

Process

1. Run cqs scout "<task_description>" --json via Bash
2. Run cqs gather "<task_description>" --json --tokens 4000 via Bash
3. Present results as an implementation brief

Output Format

Present this to the user (fill in from the JSON results):

## Investigation: <task_description>

### Relevant Files
<from scout: file groups ranked by relevance, with chunk names>

### Key Functions to Understand
<from gather: top functions with signatures, grouped by role>
- **Modify targets:** <functions that likely need changes>
- **Dependencies:** <functions called by the targets>
- **Tests:** <existing tests that cover these functions>

### Staleness
<from scout: any stale files that need reindexing>

### Notes
<from scout: any relevant project notes>

### Suggested Approach
Based on the code structure:
1. Start with <most relevant file/function>
2. Check <dependency> for the interface contract
3. Test coverage: <good/partial/none> — <recommendation>

Rules

• If scout returns no results, try a broader query or different phrasing
• If gather returns no code (token budget exhausted), increase --tokens or narrow the query
• Always present the "Suggested Approach" — synthesize the scout+gather results into a 3-step plan
• If relevant notes mention warnings or gotchas, highlight them prominently
• If stale files are found, suggest cqs index before proceeding