skills.rest
mikalv
mikalvCOMMUNITY
Explore all skills by this author →
Website𝕏@mikalvGitHub

stress-interview

Runs verifier, reviewer, and challenger subagents in parallel to pressure-test changes before release.

Updated Aug 2, 2026
Search Tags:#code review#verification#multi-agent#subagent#adversarial validation#release readiness#parallel review
One-click install
npx skills add https://github.com/mikalv/pi-extensions --skill stress-interview-mikalv
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: stress-interview
Source: https://github.com/mikalv/pi-extensions/tree/main/packages/pi-subagent-fleet/seeds/skills/stress-interview
Command: npx skills add https://github.com/mikalv/pi-extensions --skill stress-interview-mikalv

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Single-reviewer code reviews often miss defects due to individual bias and limited perspective. This Skill coordinates three parallel review agents to pressure-test a change from verification, code quality, and adversarial risk angles before release. ## Core Features & Use Cases - Parallel Multi-Agent Review: Launches verifier, reviewer, and challenger subagents in one batch via the Pi subagent tool, collecting executable evidence, code-review findings, and skeptical risk questions simultaneously. - Two-Pass Mode: Separates specification compliance review (Pass 1) from code quality review (Pass 2) when requested with --2pass, classifying findings as under-built or over-built. - Structured Release Decision: Produces a severity-classified report (Must fix / Should fix / Remaining risk) with an Overall verdict of Ready, Needs changes, or Blocked. - Use Case: Before merging a critical feature branch, ask for a stress interview of the change to get test evidence, security findings, and adversarial failure scenarios compared side by side. ## Quick Start Ask the agent to run a stress interview on your current change or pull request to get a parallel multi-angle review with a release readiness verdict.

Frequently Asked Questions about stress-interview

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a multi-agent code review before release?▼

Invoke the stress-interview skill with your change as the argument. It launches verifier, reviewer, and challenger subagents in one parallel batch, then compares their findings into a severity-classified release decision.

What is the difference between verifier, reviewer, and challenger agents?▼

The verifier produces executable evidence like tests, type checks, and builds. The reviewer finds correctness, regression, security, and maintainability issues. The challenger raises skeptical questions about assumptions and failure scenarios.

How does two-pass review mode work?▼

Pass 1 checks specification compliance, classifying findings as under-built or over-built. Pass 2 covers code quality including correctness, security, and failure scenarios, and is re-run after critical fixes.

When should I use --isolated instead of --main for subagents?▼

Use --isolated when the review tasks are fully self-contained and should not inherit the current conversation context. Use --main when the subagents benefit from the existing conversation state.

What are the limitations of parallel multi-agent review?▼

Challenger questions are hypotheses unless backed by evidence, and conflicting agent conclusions require human judgment. The final decision does not claim certainty beyond what the collected evidence supports.

Related Skills

Explore More Software Engineering Skills→
dmmulroydmmulroy
345

code-review

Review TypeScript code changes against local coding standards with evidence-backed findings.

Community
Advanced
samuelcastrosamuelcastro

super-review

Review pull requests for merge readiness with severity-classified findings.

Community
Intermediate
thisisqubikathisisqubika

security-review

Analyze code repositories for vulnerabilities using SAST and AI triage.

Official
Advanced
try-workstry-works
4

rlm-subagent

Coordinate parallel RLM sub-phase execution with sequential fallback and two-stage reviews.

Community
Advanced