Quality

Skill Forge Eval runs evaluation pipelines for Claude Code skills to check triggering accuracy, workflow correctness, and output quality. It creates eval metadata, runs executor and grading agents, and produces benchmark and feedback reports.

Vibe Coding Harness checks other vibe-coding-kit outputs against harness.json and CLAUDE.md. It verifies required files, sections, fields, cross-skill consistency, and workflow progress before delivery.

B-Roll Runtime Planner reads and verifies legacy runtime planner handoffs from pre-v0.9 production records. It checks plan JSON, shot coverage, and preserved evidence without converting the record into a v1 workflow.

Broll Master Integrate reads legacy HyperFrames Integrator records for pre-v0.9 recovery tasks. It verifies block order, timing, asset consistency, and reports defects without modifying the old task.

Broll Remotion Render reads legacy Remotion render and delivery records for pre-v0.9 tasks. It verifies identity, hashes, media facts, and evidence when recovering old production records.

Broll Render checks legacy HyperFrames render records for pre-v0.9 tasks and verifies that recorded artifacts, hashes, and approvals match the same source closure. It is used only for recovery and audit, not for new v1 production runs.

Brownfield Drift checks code against PLAN.md boundaries to find architecture drift, cross-service imports, and circular dependencies. Use it in PR reviews or scheduled health checks when you need to verify a change follows the defined architecture.

Debugging helps reproduce an issue, inspect logs and errors, isolate the cause, then fix and verify the result. Use it when troubleshooting failures in code or tooling.

Legacy Hybrid Integrator checks frozen hybrid integration records for older pre-v0.9 tasks. It verifies planned blocks, assembly order, preview facts, and identity records without rerunning or repairing the task.

Legacy Hybrid Render reads and verifies frozen records for pre-v0.9 hybrid render tasks. Use it to inspect old approvals, receipts, and final media evidence without rerunning delivery or modifying artifacts.

Legacy Remotion Integrator checks archived pre-v0.9 Remotion task records and verifies the recorded project, assets, and identity. Use it when recovering an old task; do not use it for new v1 production work.

Module Audit Agent audits a module’s purpose, dependencies, risk, and test coverage to decide whether it is safe to refactor. It uses graph-based analysis and only generates tests when coverage is low and risk is elevated.