xenitv1/claude-code-maestro
Overview
This skill enforces a strict verification-first workflow: never claim success without fresh, reproducible evidence. It treats any unverified completion statement as unacceptable and requires running concrete commands and capturing their outputs before declaring a task done. The goal is to eliminate assumptions, prevent false positives, and restore trust in reported outcomes.
How this skill works
Before any positive status or satisfaction is expressed, the skill identifies the exact command that proves the claim, runs that command in this message, and inspects full output and exit codes. If output confirms the claim, the skill reports the claim accompanied by the captured evidence (logs, test summaries, diffs). If not, it reports the real status with the evidence and avoids any misleading language.
When to use it
- Before saying tests, linting, build, or type checks have passed
- Before marking a bug fixed or a regression resolved
- Before committing, creating a PR, or signaling task completion
- When delegating work to another agent or accepting agent-reported success
- When moving to the next task or reporting readiness for review
Best practices
- Always run the definitive verification command (npm test, npm run build, git diff, etc.) in the same context as the claim
- Capture full outputs and exit codes; include exact command invocations in reports
- Adopt red-green cycles for regressions: run tests before and after changes to prove fix
- Avoid subjective words (should, probably, seems); use only evidence-backed statements
- Fail fast and report failures with logs and reproduction steps instead of optimistic claims
Example use cases
- Run npm test and attach the 0-failures output before declaring ‘All tests pass.’
- Execute npm run build and include the compiler output prior to claiming a successful build.
- Check git diff and git status before accepting an agent’s ‘completed’ message and include the diff in the report.
- Run linter and type checks and paste the exact tool output before stating ‘Lint clean’ or ‘Type check OK.’
- Perform a red-green regression sequence (write test, run to fail, fix, run to pass) and provide each run’s output as evidence
FAQ
Sufficient evidence is the fresh, complete output of the authoritative command(s) and exit codes or diffs that directly prove the claim.
Can I rely on previous runs?
No. Only evidence generated in the current verification step supports a claim; past runs are not acceptable.
10 skills
This skill enforces verification before completion by generating fresh evidence like logs and test outputs to prove success.
This skill surgically fixes root causes in JavaScript code by enforcing forensic debugging, reflection, and architecture-aware fixes.
This skill enforces strict test-first discipline to deliver robust JavaScript code by guiding you through the red-green-refactor cycle.
This skill helps you implement pixel-perfect frontend experiences that align with human-centered psychology and production-grade design systems.
This skill helps you optimize frontend and backend performance by enforcing INP under 200ms, partial hydration, UUIDv7 indexing, and token stewardship.
This skill helps turn ideas into concrete specs by guiding design-first brainstorming, clarifying requirements, and presenting iterative, testable plans.
This skill enforces readable, secure JavaScript code by applying the clean-code gates, preventing over-engineering and hallucination-prone outputs.
This skill helps you craft Manifest V3 browser extensions with robust persistence, security, and cross-browser compatibility.
This skill creates isolated git worktrees for feature development with safe directory selection, validation against .gitignore, and baseline tests.
This skill acts as an Elite Software Architect to govern complex repositories, enforcing Why over How and maintaining memory for cross-session continuity.
