A second, read-only loop that babysits a first one. Every ten minutes it reads the run's state file and latest log, runs the test suite so its evidence is an exit code rather than a self-report, and applies four explicit escalation triggers (no progress across two checks, repeated identical stack trace, budget exceeded, merge-queue conflict). On any trigger it prints ESCALATE and stops so a human looks; it never restarts, edits, merges, or pushes. Exits when the state file reports PASS or DONE, and after twelve checks regardless. Adapted from ECC's loop-operator agent (affaan-m/ECC, MIT), which lists the same triggers but has no cadence, cap, check command, or exit.
/loop every 10 minutes act as the loop operator for the autonomous run in this repo: run `cat state/loops/*.md`, `tail -100 logs/latest.log`, and `npm test` each pass, then report the active loop pattern, current phase, last successful checkpoint, the npm test exit code, failing checks, and cost drift against the budget in the state file. Print ESCALATE and stop this loop if any of these is true: no progress across two consecutive checks, the same stack trace repeats, cost exceeds the budget, or a merge conflict is blocking the queue — escalate to me, never fix it yourself. Otherwise print CONTINUE with one line of evidence. Report-only: never edit code, never restart or kill processes, never merge or push — any PR the run opens waits for me. Exit when the state file says Status: PASS or DONE. Stop after 12 checks regardless.
Write a one-page brief where every claim has three or more sources and every link is opened and confirmed to support the claim — the loop that catches hallucinated citations a single prompt never can.
/goal write a one-page brief on [TOPIC] where every claim has at least three sources and every link opens to a real page that supports the claim. Open each link to confirm it before you call it done. Replace any source that is dead or does not back up the claim. Done when every source check passes on every claim — maximum 30 iterations.
/loop cadence: on-demand queue. For each incoming question in the queue, a cheap model attempts an answer and self-checks against [CRITERIA]. Append (question, cheap-model verdict, pass/fail) to state-file escalation-log.md. Escalate to Fable ONLY where the cheap model logged a failure; Fable answers just those. Read + own-file writes only. Stop when the queue is empty; log errors and stop. Budget: cheap-first, cap $[X]/run. Hard cap: stop after 1 iteration per run.
Reduce the CSS styling code [site] sends to users without changing tested screens. First capture representative pages, sizes, themes, and interactions, and record the built CSS size. Treat coverage reports only as suggestions. Remove one declaration or rule, rebuild, and rerun screenshots and project checks. Keep it only if every screenshot is pixel-identical and built CSS is smaller; otherwise revert. Stop when no supported candidate remains, progress stalls, or approval is required. Return reduction, evidence, and untested states.