UB Governance

Source: .agents/skills/ub-governance/SKILL.md

ub-governance is the control skill. It decides how much evidence, testing scrutiny, exception handling, ADR alignment, release discipline, or gate structure a change needs.

For the lean versus advanced model, read the UB Governance deep dive.

Core Principles

• Default to the lean profile unless advanced governance is explicitly justified.
• Keep gate states deterministic: pass, fail, or blocked.
• Treat ordinary Level 1 workflow-backed work as satisfied by workflow artifacts unless risk or scope requires more.
• Escalate ADR and claim machinery only for durable, repository-wide, high-risk, or explicitly governed decisions.
• Keep bounded exceptions owned, time-limited, and follow-up driven.
• For behavior-changing work, prefer behavior-first TDD: red, green, refactor, with regression-first proof for reported defects.
• Use testing-signal review to reject low-value tests without making every realism concern a hard blocker.

Behavior In Practice

• Selects a governance mode first: repository controls, testing posture, evidence decisions, core contract semantics, or a full governance audit.
• Defaults to lean. Advanced governance is an explicit escalation, not a background tax on ordinary work.
• Uses workflow artifacts as the durable record for ordinary Level 1 work, then escalates to ADR or claim machinery only for repository-wide, high-risk, Level 2, or explicitly governed decisions.
• Applies deterministic gate language: pass, fail, or blocked, with traceable artifact paths rather than vague approval language.
• Reviews test quality with named signals. Type redundancy, interaction-only assertions, pass-through tests, and happy-path-only suites are blocking signals; internal-detail bias is warning-only; functional-realism concerns scale with risk.
• Pushes tests toward DAMP readability over over-DRY helper abstraction when duplication makes scenario setup, trigger, and observed outcome easier to understand.
• Uses the Prove-It defect pattern: reproduce the reported defect with a failing regression test, confirm the failure path, make the smallest fix, then rerun narrow and broader validation.
• Checks RED realism before implementation. A red test is not useful if it mocks the behavior under review, encodes the answer in a fake, or would pass if the real implementation disappeared.
• Keeps exceptions bounded with owner, rationale, creation date, expiration date, and follow-up. An exception is not an informal permission slip.
• Distinguishes repository governance from this repository's maintenance scripts. Skill catalog checks, README drift checks, and local path integrity are factory controls, not portable governance by default.

Reference Highlights

• .agents/skills/ub-governance/references/profile-model.md: lean and advanced governance profiles, escalation triggers, and default expectations.
• .agents/skills/ub-governance/references/testing-policy-and-signals.md: behavior-first TDD posture, blocking test anti-patterns, warning signals, functional-realism guidance, suite-balance rules, and flake policy.
• .agents/skills/ub-governance/references/decision-memory-and-claims.md: when workflow artifacts are enough, when ADR escalation is justified, and when claim-register evidence becomes part of a blocking rationale.
• .agents/skills/ub-governance/references/evidence-baseline.md: evidence expectations for higher-risk paths, validation records, and freshness.
• .agents/skills/ub-governance/references/gate-and-report-contract.md: canonical gate types, gate states, report sections, and domain extensions.
• .agents/skills/ub-governance/references/exception-contract.md: bounded exception metadata, common exception templates, expiration requirements, and follow-up rules.

Progressive Disclosure

The main skill handles ordinary governance routing. Deeper references load by mode and risk: testing policy for test-signal review, evidence references for Level 2 or high-risk work, repository references for repository controls, and schema files only when concrete governance data is being authored or checked.

Common Invocation Examples

• “Use ub-governance to decide whether this needs an ADR.”
• “Check whether this feature should follow red/green/refactor.”
• “Review this test plan for low-signal testing patterns.”
• “Tell me if this test is too DRY and hides the behavior.”
• “Check whether this exception is bounded enough.”
• “Explain the lean versus advanced governance choice.”
• “Does this change need claim evidence or are workflow artifacts enough?”

Boundaries

Do not use it as a general workflow planner. Use ub-workflow to choose delivery lanes, PRD shape, sprint flow, and stop-resume behavior.

Tradeoffs

Strength: keeps ordinary work lightweight while preserving escalation paths for risk, auditability, and durable decisions.

Cost: advanced governance adds overhead and should be activated deliberately.

Deep Dive

See UB Governance deep dive for profile selection, governance modes, gate states, and bounded exceptions.