Skip to content
cccskills
FreePublished on GitHub

pui-unclaim

Release one current Proto UI work-item claim when it expired, became invalid, or the contributor stopped. Use only for the contributor's own recorded claim and only when the exact reversible GitHub mutation is authorized. Do not select replacement work, change readiness, or release another contributor's claim.

The English summary is being prepared. Showing the original description.

Works withGitHub
See how to install

Files included (2)

  • SKILL.md3.9 KB
  • agents/openai.yaml246 B

SKILL.md (original)

Review the instructions the agent will receive before you install.

Release one claim

Owner-authorized dot exception

The owner-authorized dot exemption in internal/agent-operations/contributor-agents.md takes precedence over this leaf's ModelTrace-specific sampling, record/context and disclosure requirements. When acting as dot, do not sample or invent a ModelTrace record: visibly identify Agent: dot and the exact not-measured exemption declaration at every write. Use the supported publisher's explicit dot branch or the guide's real connected-service path; never fake a measured artifact, a successful validator, or live gh responses. The exemption grants no action permission, model authentication, independent approval or repository-rule bypass. All non-ModelTrace inputs and existing authorization, exact-head/tree, DCO, CI, independent-review and readback requirements remain. Other Agents follow the measured path below unchanged.

  1. Require a current pui-orient envelope. In autonomous mode the release must be within the fresh C2-or-higher ceiling; in human-assisted mode the assessment remains advisory. Require the active context's current content-bound modeltrace-record in both modes. Validate it against independent context and include its generated public receipt in the authorized release notice. Missing, expired or changed-scope records route back through the entrypoint for remeasurement, never declared-label fallback.
  2. Read the original claim receipt, current issue, recent comments, linked work, assignee, and Project claim state when available.
  3. Require one explicit release reason: expiry, changed boundary, invalidated task state, blocking dependency, stopped work, or completed handoff.
  4. Revalidate live GitHub permission, current authorization, claim ownership, target version, and the idempotency key for the requested release action.
  5. Post exactly one request-bound release notice and clear only claim metadata that the current contributor owns and is authorized to change.
  6. After a timeout or other unknown write outcome, reconcile the exact claim and release marker once. Attribute success only when the intended owned metadata is released; otherwise return the reconciled terminal result and never retry blindly.
  7. Return a mutation receipt containing the claim identity, release reason, observed pre-state, resulting state, timestamp, and any bounded handoff.

An ownership, permission, authorization, or live-state mismatch returns an exact no-action receipt for recollection. The release boundary preserves readiness, semantics, labels, milestone, every other contributor's assignment, and implementation scope.

Publish the release notice through agent:publish -- comment with --record, --context, independently established --mode/--mode-source, and exact --authorization; supported metadata changes use separate record-aware exact-target agent:collaborate actions. Carry the private record's exact reference/digest in the handoff, not samples or private context. ModelTrace is unsigned closed-set attribution, not authentication or authority. Raw gh/API use is no policy exemption; preserve human original text and all ambiguous/failed results and anomalies.

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.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Implement one governed Proto UI Adapter-profile realization slice. Use when target and runtime range, supported or omitted Module role, Host Capability fidelity, lifecycle translation, failure behavior, compatibility boundary, and evidence scope are fixed by current authority. Use pui-adapter-assess for research or readiness assessment and pui-regression first for a governed failing behavior.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

Assess one bounded Proto UI Adapter question without changing repository or external state. Use for new-Adapter research, target compatibility, catalog readiness, reviewed Module support or omission evidence, Host Capability fidelity, lifecycle translation, dependency risk, or implementation readiness. Do not catalog, implement, repair, or approve the Adapter slice.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

Measure the active Agent model with pinned offline ModelTrace sampling and return an unsigned closed-set attribution record. Use before Agent-originated commits, Issue/PR creation or material updates, comments, reviews, and current collaboration writes in either execution mode. Do not treat a fingerprint, declared model label, or record reference as authenticated identity, permission, independent review, or acceptance.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

Run Proto UI's dynamic comprehension assessment and derive an unsigned local task-fit result. Use when an Agent needs review calibration or a fresh autonomous task and review ceiling. Do not use a self-result as identity, permission, or acceptance evidence.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

Investigate a bounded Proto UI Contract, Prototype, Module, Host Capability, Adapter, schema, or ownership question and turn a genuinely unresolved direction into one semantic decision packet. Use for spec brainstorming, admission questions, competing semantic models, uncataloged behavior, or unresolved cross-layer boundaries.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

pui-ci

Free

Diagnose one bounded Proto UI Actions or CI failure from an exact workflow run and revision, then route the exact repair, validation, or recheck transition. Use to locate the owning failure and distinguish infrastructure from change failures. This leaf remains read-only.

The English summary is being prepared. Showing the original description.

Proto-UI/Proto-UI36Updated Oct 22, 2026

See all skills by Proto-UI

Report a problem with this skill