Propose one eligible work item

1. Require a current pui-orient envelope.
2. Read CONTRIBUTING.md and inspect the live issue queue, assignees, recent comments, linked pull requests, labels, milestones, and Project fields when available.
3. Reject work that is unassessed, ambiguous, design-blocked, oversized, missing acceptance or validation boundaries, beyond capability or permission, already owned, actively claimed, or linked to active implementation.
4. Select at most one smallest ready item with a fixed expected result, explicit exclusions, evidence boundary, risk, required capability, and permission ceiling. Returning no eligible item is correct.
5. In autonomous mode, reject any item whose task class exceeds the fresh self-assessed ceiling. For review work, also require one declared review class present in recommendedReviewClasses. In human-assisted mode, report low confidence and stronger review needs without blocking the user's selected work.
6. For the selected Issue, inspect its existing evidence using internal/agent-operations/visual-evidence.md. Return an evidence-assessment artifact bound to the Issue URL, observed update time and inspected discussion cutoff. Include a sanitized request paraphrase with actual source attribution, existing verified evidence links and quality, inspected versus unknown discussion scope, baseline/result, complete/partial/blocked disposition for the named evidence scope, and each gap's reason and next Agent action. A future test plan is not current evidence. If no item is eligible, record a no-target assessment without inventing a request or defect.
7. Return raw selection facts, a copy-ready claim proposal, planned evidence, expiry, unknowns, and the exact authorization and live permission required by pui-claim. Embed the assessed Agent evidence section in the proposed claim text before requesting authorization for that exact text. Missing evidence is Agent debt, not grounds to reject human intake or an otherwise ready item.

Remain read-only. Do not comment, label, assign, claim, edit, close, create a branch, or mutate Project state. A proposal never changes readiness or grants implementation authority.

Explicit handoff

Do not load or execute another skill. Return exactly one handoff conforming to internal/agent-operations/schemas/skill-handoff.schema.json. Carry required prior artifacts by reference, include every artifact this leaf produces according to skills.yaml, and set nextSkillId to one eligible registered leaf or null.

Communicate with the user in the user's current language. Keep GitHub identifiers canonical.