Engineering Pipeline

TIP

Quick reference: Engineering workflows are grouped under src/cli/commands/engineering.ts and connect to sprint state, browse daemon, and quality gates.

Workflow Building Blocks

Browse automation for UI checks and visual captures
Guardian checks for destructive-command and path safety controls
Sprint pipeline with .cm/sprint artifacts and event progression
Advisory loop for reviewing execution analyses, skill quality, and self-healing handoffs
Second opinion / retro / suggest helpers for analysis and follow-up

Sprint Runtime Artifacts

Sprint-related runtime state is managed in .cm/sprint and linked by utilities in src/sprint-pipeline.ts.

Recommended Flow

Prepare context and sprint state.
Run implementation tasks and validations.
Run QA (including visual QA where needed).
Review and canary before broader rollout.
When a run degrades or a skill looks weak, use the advisory loop before changing skill behavior.

Advisory Workflow

Use the advisory commands when you want telemetry-backed follow-through instead of guesswork:

bash

cm advisory report --project .
cm advisory metrics --project .
cm advisory handoff --project . --for cm-skill-health

Then use the structured note with:

the legacy cm-skill-health advisory consumer id for diagnosis payloads
cm-skill-evolution for FIX, DERIVED, or CAPTURED

Risk Reduction

Gate with npm run test:gate before major transitions.
Use guardian checks for sensitive operations.
Keep sprint artifacts updated for reproducible handoffs.
Use cm advisory handoff --format json when another agent needs the same evidence without re-querying SQLite.

Engineering Pipeline ​

Workflow Building Blocks ​

Sprint Runtime Artifacts ​

Recommended Flow ​

Advisory Workflow ​