tickets-refine

Refine open tickets

Do not start refinement in the invocation turn. Return only this copyable block, with the invocation argument preserved as written:

/goal Follow the loaded tickets-refine workflow. Scope: "$ARGUMENTS"; an empty scope means every eligible ticket in open/observations and open/needs-scope-review. Continue one ticket at a time until every eligible ticket has been verified, atomized, checked for meaningful duplicates, scoped across its known blast radius, and moved to ready with an evidenced execution route, needs-decision, or archive/rejected; or stop when the user asks to pause or cancel, or with an evidenced external blocker that prevents any eligible work from continuing. Before finishing, surface the checked queue, ticket outcomes, execution routes, unresolved decision boundaries, and completion evidence in the conversation.

When that goal starts, first read ticket-autonomous-runs.md and ticket-format.md.

Work from open/observations and open/needs-scope-review. A plain-language argument may narrow the queue by ticket id, path, component, or concern but grants no new authority. For each ticket:

1. Restate a falsifiable problem and verify it against current repository evidence and, where relevant, current authoritative external contracts.
2. Find affected locations and meaningful consequences far enough to establish the blast radius. Separate independent problems into separate tickets while preserving the original history and links.
3. Compare plausible semantic duplicates, not only matching words. Keep the clearest canonical open ticket; update it with material evidence, then move confirmed duplicates to archive/rejected with a link and reason.
4. Move a disproved or superseded ticket to archive/rejected. Move work to ready with execution: autonomous only when current cited evidence fixes the expected behavior and no product, business-logic, material infrastructure, destructive-action, data, authority, or irreducible preference choice remains. Add the required autonomous boundary section. Move every genuine or uncertain user-owned choice to needs-decision with the exact decision stated plainly. Never infer autonomous eligibility merely from an existing ready path or confidence in a proposed solution.

Do not run tests, servers, containers, migrations, benchmarks, or external environments, and do not implement a fix. Existing tests and saved outputs may be read as evidence. Do not invent certainty when a claim requires a new measurement: preserve the missing evidence and route the ticket according to the actual decision boundary rather than guessing.