Release: Learning Quality And Workflow Capture

Forgeflow 4.3.40 tightens the learning path with workflow-ending capture prompts, telemetry-quality checks, command-wrapper cleanup planning, and safer evidence handling for local outcome records.

Workflow Capture

• New /forgeflow-workflow-ending-capture recommends the one recorder command to consider after review, next-work, or agent-feedback workflow endings.
• /forgeflow-outcome-capture-plan now trusts local JSONL evidence over stale project-intelligence metadata, so missing, empty, or invalid local evidence prompts capture.
• Workflow-ending capture inherits the same invalid-evidence behavior and does not suppress recorder prompts for malformed local outcome files.

Telemetry Quality

• New /forgeflow-telemetry-quality summarizes local metrics, review outcomes, agent feedback, and next-work outcomes.
• Malformed outcome or metrics lines now downgrade readiness to attention instead of reporting calibration-ready evidence.
• The helper stays local and advisory: it does not backfill, infer, export, edit files, commit, or push.

Wrapper And Profile Readiness

• New /forgeflow-command-wrapper-batch ranks the next small command-wrapper cleanup batch without editing command files.
• New command wrappers fail closed on unsupported or malformed arguments instead of silently using defaults.
• /forgeflow-profile-bootstrap now returns a machine-readable next profile action for prompt-needed, preview, and written states.

Context Budget Follow-Through

• /forgeflow-context-wave-build now reports post-build budget status and violation counts after creating a focused packet.
• Install, update, health, release, README, and wiki surfaces include the new helpers and regression coverage.

Validation

• Full scripts/forgeflow/test-*.js suite
• node scripts/forgeflow/smoke-check.js --mode source --json
• node scripts/forgeflow/test-doc-links.js
• git diff --check
• Forgeflow review approvals from Smith, Warden, Lumen, and Atlas after fixes