Prompt Examples

Ready-to-paste agent configuration blocks for deep codebase analysis workflows.

Hotspot & Risk Analysis

## Hotspot Detection

When investigating code quality or stability:
1. Broad scan: `semantic_search` with `rerank: "hotspots"`, `metaOnly: true`, `limit: 20`
2. Drill down: for top results, use `semantic_search` with `pathPattern` targeting that area

### Threshold Decision Table

| Signal           | Safe  | Caution | Stop     |
|------------------|-------|---------|----------|
| chunkBugFixRate  | < 15% | 15–40%  | > 40%    |
| chunkCommitCount | 1–3   | 4–7     | > 8      |
| churnVolatility  | < 5   | 5–20    | > 20     |

Action: Critical hotspots need redesign, not patches.

Ownership & Knowledge Silo Detection

## Ownership Analysis

To find knowledge silos:
1. `semantic_search` with `rerank: "ownership"`, `metaOnly: true`
2. Cross-reference ownership with hotspots for risk assessment

### Ownership Decision Table

| dominantAuthorPct | Status           | Action                              |
|-------------------|------------------|-------------------------------------|
| < 70%             | Well-distributed | Safe to modify freely               |
| 70–85%            | Partial silo     | Match dominant author's style       |
| > 85%             | Knowledge silo   | Flag owner for review before change |

### Risk Cross-Reference

| Silo? | Hotspot? | Risk   | Action                          |
|-------|----------|--------|---------------------------------|
| No    | No       | Low    | Safe to modify                  |
| No    | Yes      | Medium | Add tests, careful review       |
| Yes   | No       | Medium | Notify owner, match their style |
| Yes   | Yes      | High   | Escalate — redesign discussion  |

Tech Debt Assessment

## Tech Debt Discovery

To find legacy code that needs attention:
1. `semantic_search` with `rerank: "techDebt"`, `metaOnly: true`, `limit: 30`
2. Filter for confirmed debt: `git.ageDays >= 90` AND `git.commitCount >= 5`
3. Cross with hotspots: `semantic_search` with `rerank: "hotspots"`, same pathPattern

### Severity Decision Table

| chunkAgeDays | chunkBugFixRate | chunkCommitCount | Severity |
|--------------|-----------------|------------------|----------|
| > 90         | > 40%           | > 10             | HIGH     |
| > 60         | > 25%           | any              | MEDIUM   |
| > 90         | any             | < 3              | LOW      |

Blast Radius Estimation

## Blast Radius Check

Before modifying a module:
1. `semantic_search` with `rerank: "impactAnalysis"`, `metaOnly: true`
2. Count imports in results — high imports = many dependents
3. Overlay risk: `semantic_search` with `rerank: "hotspots"` on dependent files

### Risk Decision Table

| Dependents | Hotspot? | Action                         |
|------------|----------|--------------------------------|
| < 5        | No       | Safe to modify                 |
| < 5        | Yes      | Modify with tests              |
| 5–15       | No       | Careful, run integration tests |
| 5–15       | Yes      | Needs design review            |
| > 15       | Any      | Propose RFC before changing    |

Combined Multi-Step Analysis

## Deep Analysis Workflow

For thorough codebase assessment. All steps use `semantic_search` with `metaOnly: true`, `limit: 20`.

### Analysis Steps

| Step | Action                 | Rerank preset    |
|------|------------------------|------------------|
| 1    | Discover relevant code | `relevance`      |
| 2    | Assess risk            | `hotspots`       |
| 3    | Map ownership          | `ownership`      |
| 4    | Estimate blast radius  | `impactAnalysis` |

Synthesize: code that appears in hotspots AND silos AND has high blast radius is highest priority.

tip

Combine these prompt blocks with the reranking presets from Search Strategies for maximum effectiveness. The presets handle the scoring math; these workflows tell your agent when and how to apply them.