Current Objective
Build Bossman into the human-first cockpit for fleet site truth, standards, evidence, priority, actions, and agent-safe operations.
This is the standing agreement for how Codex can keep building Bossman through bounded slices without asking Jason to approve every small step.
Build Bossman into the human-first cockpit for fleet site truth, standards, evidence, priority, actions, and agent-safe operations.
Upgrade the Sites page into the main working surface for accepted sites, with priority, repo, lifecycle, and source information.
Implement a focused slice, run preflight, commit and push when green, verify production, then continue to the next ordered slice unless blocked.
Improve the Sites surface so the 25 accepted sites are readable and useful.
Add manual priority tiers for accepted sites.
Publish or verify live /bossman-site-manifest.json files for accepted sites.
Import fleet standards and per-site applicability.
Add compliance evidence summaries to each site.
Build the action queue for standards failures, drift, and follow-ups.
Add scheduled refreshes for manifests, standards, and evidence.
Add evidence-led priority signals from traffic, leads, and quote data.
Edit Bossman code, tests, docs, and views.
Run format, static analysis, tests, and preflight checks.
Commit and push focused green slices to main.
Let Forge auto-deploy from pushed commits.
Verify production pages and agent APIs after deploy.
Use protected Bossman APIs to import evidence into staging.
Create proposals from trusted repo manifests and system-owned evidence.
Approve low-risk proposals that match accepted source contracts and do not conflict with accepted truth.
DNS, Synergy, registrar, server, billing, or provider changes.
Creating, rotating, or exposing secrets or production env values.
Deleting production data or destructively rewriting accepted truth.
Approving proposals where evidence conflicts with existing accepted truth.
Changing customer quote flows, booking flows, or live payment behavior.
Broad UI/product direction changes not covered by the ordered slices.
Failed tests, failed deploys, or unclear production behavior when the fix is not obvious and low risk.
The durable markdown version lives at docs/autonomous-execution-plan.md.