Dispatch two competing reviewers (arch-reviewer and impl-reviewer) before deep analysis. Competition produces more thorough results. Use before creating code, modifying architecture, making technical decisions, or answering codebase questions.
SkillScore59Maintained
Automatic
◆QA & Test Engineering pack
Last verified 9/3/2026
What it changes
Passed 13 of 22 test cases with the skill, and 4 of 22 without it.
Without skill18%
With skill59%
Checks it passed that the model alone missed
Dispatches arch-reviewer and impl-reviewer
Lists verified facts separate from combined issues
Spawns task for arch-reviewer with prompt parameter
and 15 more in the full report
9 fixed4 passed either way9 still failing0 broke
Pass-rate lift
+41 pp
Tokens
+21%
Agent turns
0%
Response time
-1%
Model gemini-3.6-flashMeasured 9/3/2026v113 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
Our engineering team wants to replace Celery with Temporal for orchestrating multi-step saga transactions. Run a competitive review across system design and code implementation before deep dive analysis.
What was checked, without and with the skill
Checked
Without skill
With skill
Fact-checks technology claims regarding Temporal and Celery
Passed
Passed
Includes Reviewer, Issues Found, HIGH, MED, LOW in results table
Missed
Passed
case-13 · gemini-3.6-flash · 9/3/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md123 linessingle file
Competitive Review
Dispatch two competing reviewers before deep analysis. Competition produces more thorough results.
Purpose
Different perspectives catch different issues. Architecture reviewers find structural problems;
implementation reviewers find code-level bugs and fact-check claims. Running them in competition
("whoever finds more issues gets promoted") increases thoroughness.
Triggers
Use before ANY complex task involving:
Creating new code
Modifying existing architecture
Making technical decisions
Answering questions about a codebase
Building new features
Protocol
Step 1: Announce the Competition
Say: "I'm dispatching two competing reviewers to analyze this."
Step 2: Spawn Both Agents IN PARALLEL
Task(agent="arch-reviewer", prompt="[full user question + context]")
Task(agent="impl-reviewer", prompt="[full user question + context]")
Tell each agent:
"You are competing against another agent. Whoever finds more valid issues gets promoted. Be thorough."