P1: Measure governance-plane control effectiveness and agent ergonomics after pilot adoption

[BunsDev]

Outcome

Define and collect evidence showing whether the governance plane improves coordination and agent readiness without becoming an expensive central bottleneck or a false-compliance dashboard.

This measurement begins only after PR #7 is merged, #6 applies the administrative baseline, and #8 lands at least two immutable pilot adopters.

Measures

Collect by repository, risk class, and task type:

• clean-clone bootstrap success;
• fast/full check duration and flake rate;
• clarification count before an agent can identify the canonical owner;
• human interventions for R0/R1 work;
• correctly stopped/proposed R3/R4 attempts;
• stale manifest, generated-view, dependency, owner, and GitHub-setting drift;
• false-positive and escaped-drift rate;
• reusable workflow adoption and immutable-pin freshness;
• exceptions opened, expired, renewed, and closed;
• rollback/recovery success;
• cross-repository compatibility regressions;
• ownership conflicts and time to resolve;
• contributor friction and time-to-first-green-PR;
• public/private leakage test results;
• administrator and backup-reviewer coverage.

Guardrails

• Do not use repository count, green badges, or manifest presence alone as readiness proof.
• Do not reward agents for bypassing approvals or reducing test scope.
• Keep operational data minimized; do not publish private repository names, prompts, memories, user data, personal performance profiles, or security incident detail.
• Represent unsupported, unavailable, stale, and degraded evidence explicitly.
• Separate structural validity, repository verification, administrative application, control effectiveness, and product/protocol conformance.

Acceptance criteria

• Baseline is captured before broad rollout.
• Metrics have source, calculation, owner, review cadence, retention, and privacy classification.
• At least one adversarial exercise tests metadata spoofing, reusable-workflow pin drift, and private-data leakage.
• At least one recovery exercise rebuilds generated views and governance evidence from authoritative inputs.
• A 30-day pilot review recommends continue, modify, pause, or replace with explicit evidence.
• A 90-day review decides whether file-based governance still scales or a projected service catalog is justified.
• Any future service remains a projection unless a separate ADR explicitly migrates authority with rollback.

Measurement results inform governance decisions; they do not certify security, privacy, SOC 2, ISO, GDPR, NIST, OpenSSF, SLSA, or product conformance.