hr-governance-learning

name	hr-governance-learning
description	Adaptation plane — learning ledger, reflect/promote L0-L3, DMAIC/PDCA, CTQ snapshots, router hints. Human-invoked for RGF workflows.
disable-model-invocation	true
paths	specs/governance/**,scripts/governance-learning.mjs,scripts/governance-evidence.mjs,docs/meta/governance-continuous-learning.md
minRiskTier	T2

HR governance learning (Adaptation plane)

Reflective Governance Fabric (RGF) + DMAIC/PDCA (ADR 0018, ADR 0019). Human invokes; agents do not self-promote policy.

When

• Running governance:reflect (includes ctqSnapshot)
• Opening DMAIC project or PDCA loop
• Promoting L1/L2/L3 with gates below

PDCA checklist (weekly)

• Plan: reflect report + stub
• Do: shadow router hint or L1 trial
• Check: improve measure --ctq … vs baseline
• Act: promote or reject with principal

DMAIC checklist (project)

• Define — --ctq or --brief required on improve define
• Measure — baseline + optional --ctq metric
• Analyze — RCA stub
• Improve — promote L1/L2/L3
• Control — control plan (L2 required)

Learning tiers

Tier	Mutates	Gate
L0	Ledger	Automatic
L1	Skill references/learning/*.md	PO + CI
L2	adaptation.skillRouterHints	Control plan
L3	additionalPathTriggers	Counsel + ADR + sync-check

Evidence (ADR 0019)

• T3+ handoffs: evidenceBundlePath + lane sign-offs
• Collect: npm run governance:evidence:collect
• Verify: npm run governance:evidence

Do not

• L2 without control plan; L3 without --adr
• Define DMAIC without CTQ or brief
• Weaken non-learnable invariants

References

• governance-continuous-learning.md
• product-value-ctq-tree.yaml
• ADR 0019