| name | escalate |
| description | Handle the two situations the greedy ratchet cannot: route safety-critical or high-risk changes to human review, and switch search strategy when the loop plateaus (greedy local search -> grid / tree / evolutionary). Use from inside a ratchet-experiment loop. |
Status: v0 scaffold — unverified. Behavior not yet validated end-to-end in this
repo. See ../../STATUS.md.
escalate
The keep/revert ratchet is greedy and autonomous. Two conditions require escalation.
Trigger 1 — safety-critical / high-risk change
If a proposed change touches a safety-relevant path (planner, controller, anything
that could cause real-world harm) or is otherwise high-risk:
- Stop autonomous merging. Treat the result as candidate evidence only.
- Open a human-review request with: the diff, the measured effect, the evaluator
used, and the residual risks.
- Proceed only after a human approves. See
../../AGENTS.md.
Trigger 2 — the loop has plateaued (local-search trap)
If N consecutive experiments fail to improve the metric beyond measured noise, the
greedy ratchet is likely stuck in a local optimum (see
../../docs/pitfalls.md). Switch strategy:
- Factorial grid — sweep several variants in parallel to catch interaction
effects (see
../../src/talos/orchestration.py).
- Tree search — branch from the best node, not just the latest (AIDE-style).
- Evolutionary search — a population with novelty rejection (ShinkaEvolve /
OpenEvolve-style) when single-lineage search is exhausted.
- Or hand a moonshot to a human to seed a bigger architectural bet.
Output
Either a human-review request (Trigger 1) or a switch to a broader search procedure
(Trigger 2), recorded in the ledger so the change of strategy is auditable.
