Governance
Autonomy advances only when the evidence supports it.
You get enforceable policy, review queues, branch-protection binding, spend caps, model-routing rules, PR-storm controls, and automatic autonomy demotion when quality drops.
Governance engine
Policy is enforced before the agent starts.
Boetica binds policy, branch protection, human review, budget caps, model routing, quality gates, and PR-storm controls to every build and remediation run.
preflight.policy() -> allowed / blocked / reviewbudget.hardStop() -> no runaway spendautonomy.demote() -> rollback spike detectedPolicy DSL
Rules evaluate repo, branch, environment, task kind, file path, control family, migration risk, spend, and PR volume before work starts.
Review queue
Mandatory reviews create assigned requests with SLA, escalation, decision, comment, and feedback records.
Autonomy guardrails
Repos graduate through measured autonomy levels and demote automatically on rollback, intervention, CI, or quality-gate spikes.
Policy
Rules are enforced outside the pitch deck.
The same policy gate evaluates prompt builds, backlog dispatch, schedules, initiatives, foreign-PR babysitting, remediation, and API-created runs.
- Branch-protection binding
- Protected file patterns
- Migration and dependency gates
- Human merge requirements
Budget
Spend is visible before and after every run.
Preflight estimates cost, hard caps stop runaway tasks, alerts route to notification rules, and reporting ties spend to accepted PRs and evidence packets.
- Pre-run estimate
- Hard budget stops
- Cost per accepted PR
- Budget alert delivery
Signed remediation PR
boetica/auth-rate-limit-pr
@@ apps/web/src/app/api/login/route.ts- await password.verify(body.password)+ await limiter.consume(requestIp)+ await password.verify(body.password)+ await audit.append(evidencePacket)
- packet hash
- 91f4e2a0
- prev hash
- 5e1c8a44
- trust boundary
- sandbox attested · egress denied
- merge policy
- platform-eng / human merge
Quality
Green CI is necessary, not sufficient.
Boetica also records model-route decisions, output evals, no-test justifications, diff-size controls, blast-radius caps, risky-file reviews, and abuse-control pauses.
- Model-routing policy
- Output evaluation
- PR-storm circuit breaker
- Automatic demotion
