AgentOps now describes the selected Context Delivery Lifecycle alongside standalone RPI: external reviewed memory, native work and runtime ownership, independent validation, and explicit limits on what the pilots proved. The active ADRs name which historical prescriptions are superseded and which invariants remain. This integrates the BD foundation, including doctor diagnostics that distinguish BD from BR and protection against private tracker data entering outgoing Git history. It also fixes skills audit resolution of explicit relative reference paths. Source-owned projections were regenerated, and the landed stopping behavior is preserved. Validation: Go build/vet/test and race/shuffle, 1,325 Bats cases, the local aggregate, and full AO gates. The gate registry reports 71 passes and one nonblocking preamble-adoption failure on the existing report script; that advisory remains disclosed. Historical probe receipts whose source hashes changed remain historical evidence. This is foundation adoption. It does not claim that the later memory, goal, background, privacy-release, or benefit-measurement work is implemented. Independent Astra and Fable judgments both PASS over the exact integrated content. Earlier negative results and their factual resolution remain preserved in private work evidence.
2.3 KiB
Bushido Refinery — continuous main-validation backstop
RETIRED: The refinery command and lifecycle described here are no longer live. BR and
ao beads direxamples below preserve historical behavior; they are not current tracker instructions. Use the repository operating contract and native BD for current work.
The refinery (ao refinery) is the backstop half of the push-to-main model
(ag-qidx). Push-to-main makes the local pre-push gate the pre-merge wall; the
refinery is the always-on net behind main on bushido.
What it does
Every tick it checks origin/main. On a new commit it runs the full gate
(ao gate check --full). On a blocking failure it:
- re-runs each failing check N times (default 3) to tell deterministic from flaky (the repo's 18–30% flake rate means naive escalation would be noise);
- for deterministic failures only: writes a poison beacon (
.refinery-poison- a
refinerygit note on the bad SHA) and files a blocking fix-bead (BEADS_DIR="$(ao beads dir)" br create --labels refinery,blocking);
- a
- on green: clears the beacon.
It never reverts. A poisoned commit stays on main; the team fixes forward.
Revert remains a human/quorum decision (ag-qidx P2.5, deferred to the quorum
infra ag-k99u).
Backstop, not gatekeeper
If bushido (or its Wi-Fi) is down, the refinery is simply blind — merges still
succeed, nothing blocks. On restart it resumes from .refinery-state
(last_checked_sha) and catches up. This is the deliberate posture after the
2026-06-05 control-plane crash: no single host is in the merge path.
Run it
ao refinery once # one tick (manual / cron)
ao refinery run --interval 5m # loop (the systemd service)
Install on bushido
cp deploy/agentops-refinery.service ~/.config/systemd/user/
systemctl --user daemon-reload
systemctl --user enable --now agentops-refinery
systemctl --user status agentops-refinery
journalctl --user -u agentops-refinery -f # logs
State + beacon (repo-root, gitignored runtime):
.refinery-state—{last_checked_sha, poison[]}.refinery-poison— present iffmainis currently poisoned
Tuning
--interval— poll cadence (default 5m).- Re-run count is 3 (deterministic = fails all 3). Raise for noisier suites.