The QA Acceptance skill automates the validation of software features against predefined technical benchmarks and business requirements. By evaluating pipeline runs against a structured acceptance matrix, it ensures that new releases meet critical performance metrics and stability standards through automated smoke tests and detailed quality reporting.
Use Cases
Post-Feature Release Verification: Automatically validate new features against specific technical thresholds (e.g., accuracy, latency) defined in the ACCEPTANCE_MATRIX.md to ensure quality standards are met.
CI/CD Quality Gates: Integrate strict smoke tests into automated pipelines to block builds that fail to meet core invariants or performance requirements before they reach production.
Automated Performance Reporting: Generate comprehensive QA reports that compare real-time metrics against targets, providing clear pass/fail/warning statuses for stakeholder review.
Regression Testing for ML Models: Execute targeted pytest suites for specialized components like computer vision models or TensorRT embeddings to detect regressions in accuracy or speedup.
name
qa-acceptance
description
Evaluate pipeline runs against ACCEPTANCE_MATRIX.md. Use after feature releases to verify acceptance criteria.
QA Acceptance Skill
Use this skill to validate features against acceptance criteria.
When to Use
After implementing a new feature
Before promoting from FEATURES/ to production
Running acceptance checks in CI
Generating QA reports for release
Debugging why acceptance checks failed
Key Skills
Run smoke validation against ACCEPTANCE_MATRIX.md
Use the smoke runner to validate required artifacts + basic invariants (and optionally fail if anything is skipped).
# Minimal (will mark missing prereqs as skipped)
python -m tools.smoke.smoke_run --episode-id <EP_ID> --max-frames 300
# Strict (fails if any requested stage is skipped/failed)
python -m tools.smoke.smoke_run --episode-id <EP_ID> --max-frames 300 --strict
# Include body-tracking artifact checks (requires you ran FEATURES.body_tracking separately)
python -m tools.smoke.smoke_run --episode-id <EP_ID> --body-tracking --strict