skill

Three-Musketeers

Public Unreviewed

by Drew Owens

Execute complex plans and projects with three agents — a Builder, an Adversarial Reviewer, and a Verifier — looping until it genuinely works. The build half of the plan->build pipeline.

multi-agentbuildverificationadversarialclaude-codeskill
Overview Files Skills Discover Packages Releases Dependencies Issues Breadcrumbs Pull Requests Activity
three-musketeers / triggers.test.yaml
Download
1234567891011121314151617181920212223242526272829
# Trigger test spec for skill-verifier — curated.
# Boundary under test: three-musketeers owns the BUILD/VERIFY loop. It must
# fire for "build it, have something attack it, verify it works" and must STAY
# QUIET for plan-stress-testing (adom-plan) and post-hoc audits (deep-audit).

positives:
  - "This is a big feature — build it, have a separate agent try to poke holes, then verify it actually works"
  - "Drive this long build with three agents and loop until it's genuinely done"
  - "I want one agent to build, another to adversarially review, and a third to verify as a user"
  - "Orchestrate this build with a builder, an adversary, and a verifier"
  - "Run a build loop where the builder fixes whatever the reviewer breaks until it passes"
  - "use the three musketeers approach on this build"
  - "/three-musketeers"
  - "build and verify this with multiple subagents"

negatives:
  - "stress-test my plan before I start building"
  - "vet this plan with adversarial debate"
  - "audit this finished codebase and find what we missed"
  - "review the project for anything a single builder overlooked"
  - "just confirm this one-line fix works"
  - "what is the weather today"
  - "write a python function to reverse a string"

neighbors:
  - "adom-plan"
  - "deep-audit"

trials: 4