interrogate

Files (5)

• SKILL.md
• code-quality-review.md · references/
• lead-judgment.md · references/
• reviewer-prompt.md · references/
• rubric.md · references/

SKILL.md (first lines)

Read [pstack-runtime](../pstack-runtime/SKILL.md) before executing this workflow. It defines capability checks, model defaults, and fallback behavior.


# Interrogate

Spawn one reviewer per configured model to adversarially review code changes. Each model gets the same prompt and rubric. The adversarial signal comes from model diversity, not assigned personas.

The deliverable is a synthesized verdict. Do NOT auto-apply changes.

## Step 1, Determine Scope

Identify what to review from context:

- If the user points at specific files or a diff, use that
- If on a feature branch, run `git diff main...HEAD` (or the appropriate base branch) for the full changeset
- If the user's message references recent work, gather the relevant files

Package the diff (or file contents) plus any surrounding context files the reviewers need to understand the code.

## Step 2, State the Intent

Before spawning reviewers, state the intent explicitly. Derive this from:

- The user's message
- Commit messages
- PR description if one exists
- The code itself

Write one clear paragraph. If you're unsure about the intent, ask the user before proceeding.

## Step 3, Spawn Reviewers

Launch all reviewers in a single message using the harness delegation capability. Use the `interrogate reviewers` list from `.pstack/models.md` when present, one reviewer per entry, extending or shrinking the Reviewer A/B/C/D labels below to the configured entry count. Otherwise use the table defaults.

| Subagent | Default model |
|----------|---------------|
| Reviewer A | `inherit-parent` |
| Reviewer B | `inherit-parent` |
| Reviewer C | `inherit-parent` |
| Reviewer D | `inherit-parent` |

For each reviewer:
- Role: general-purpose worker; pass the referenced prompt through the available delegation API.
- `model`: the configured `interrogate reviewers` entry, or the table default with no configured line
- Permissions: read-only; use only read operations, including connector reads.

If an explicit model is unavailable, disclose the gap and use the current model if the user's constraints permit. Otherwise leave that reviewer unrun. Do not modify configuration or open a separate PR automatically. For `inherit-parent` or `auto`, omit the model override.

Read `references/reviewer-prompt.md` and fill in the template with:
1. The stated intent
2. The diff or file contents
3. The review rubric from `references/rubric.md`
4. The code-quality lens from `references/code-quality-review.md`

The same filled template goes to all reviewers, so every model applies the code-quality lens.

## Step 4, Synthesize

As results come back, build a unified picture:

1. **Parse all findings** from the reviewers. Record whether these were independent agents/models or sequential passes by one agent.
2. **Identify corroboration** only when separate reviewers independently supplied it. In same-agent fallback, repeated findings are repeated observations, not independent votes, consensus evidence or increased confidence. Ground confidence in reproduction, source and tests.
3. **Identify lone-model findings**. Still worth reading, but weight accordingly.
4. **Deduplicate**. Different models may describe the same issue differently. Merge these and note which models raised it.
5. **Note disagreements**. Distinguish disagreements between independent reviewers from alternative hypotheses considered by the same agent.

## Step 5, Lead Judgment

You are the lead reviewer, a pragmatic senior engineer, not a neutral aggregator.

Read `references/lead-judgment.md` for the full framework.

Categorize every finding using these buckets:

- **Act on**. Real issues affecting correctness, security, or maintainability given the actual goals. These would block a real PR.
- **Consider**. Legitimate points, but you're not sure they outweigh the cost of addressing them right now. Worth the user's attention.
- **Noted**. Technically valid but not actionable. Context-dependent, premature optimization, or low-impact given the current stage.
- **Dismissed**. Wrong, nitpicky, or missing context. Brief explanation why.

Before presenting findings, verify each file/line citation against the complete on-disk file with line numbers (for example, `nl -ba <file>`). Do not derive source line numbers from an excerpt with the header or blank lines removed. If exact locations cannot be checked, cite the symbol without inventing a line number.

For each finding, include:
- Which model(s) raised it
- The category (act on / consider / noted / dismissed)
- A one-line rationale for the categorization

## Output Format

Present the verdict in this structure:

### Intent
> [The stated intent paragraph from Step 2]

### Reviewers
- Reviewer [label]: [model name], [N findings] (one bullet per reviewer)

### Act On
[Findings that should be addressed. For each: description, which models raised it, why it matters.]

### Consider
[Findings worth thinking about. For each: description, which models raised it, tradeoff involved.]

### Noted
[Valid but low-priority. Brief list.]

### Dismissed
[Rejected findings with brief rationale.]

### Agreement Map
[For independent reviewers, where did they agree or diverge? For same-agent passes, name repeated observations and changed hypotheses without vote counting or confidence inflation. State that repetition provides no independent corroboration.]

From demouo/pstack-general, MIT.

More from demouo/pstack-general