harden-task-file
Harden a manifest-dev task guidance file for one-shot quality — either /define''s quality-gate/Default set or figure-out''s probe set. Iterates: orthogonality…
Author self-sufficient Tickets in the project''s store from a finished Manifest, a direct work request, an open question, or follow-up findings from another Ticket. A Manifest becomes one coherent Shaped Ticket by default; split it by Deliverable only when the caller explicitly
$ npx -y skills add doodledood/manifest-dev --skill ticket-up --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/ticket-upContext preview
The summary Claude sees to decide when to auto-load this skill.
Author self-sufficient Tickets in the project''s store from a finished Manifest, a direct work request, an open question, or follow-up findings from another Ticket. A Manifest becomes one coherent Shaped Ticket by default; split it by Deliverable only when the caller explicitly
name: ticket-up description: 'Author self-sufficient Tickets in the project''s store from a finished Manifest, a direct work request, an open question, or follow-up findings from another Ticket. A Manifest becomes one coherent Shaped Ticket by default; split it by Deliverable only when the caller explicitly wants delegation or parallel pickup. Use when creating Tickets, ticketing up a plan, recording follow-up work, or handing work to a teammate, agent, or later session.'
`ticket-up` is the single Ticket-authoring boundary. It shapes prose, applies the convention, deduplicates, and renders the result through the project's configured venue.
Read `references/TICKET_CONVENTION.md` before emitting. It defines the unit, kinds, anatomy, lifecycle, and priority every venue must preserve.
Accept any of these inputs:
Without explicit input, use a recent Manifest only when the conversation already establishes it as the intended source; otherwise ask what should become a Ticket.
One Ticket represents one independently schedulable lifecycle. Bundle work that shares one outcome and would be assigned, prioritized, and closed together. Split only where separate ownership, priority, blocking, or closure has real value. Do not turn every Deliverable, question, or minor finding into a Ticket.
Whether to split is one question; where the cut falls is another, and the second is answered the same way whatever the input was. Every Shaped Ticket a split produces is a slice someone can exercise end-to-end on its own — run, read, or otherwise judged in the situation it is for, not merely inspected as present. A Ticket cut along an implementation layer ("the schema", "the endpoints", "the screen") has nothing to run, so its definition of done falls back to checking that something exists and whoever picks it up cannot tell whether the work helped. Re-cut work that arrives framed in layers instead of passing the layers through; renaming a layer does not re-cut it. A Question Ticket is outside this — an investigation has no slice to run, and its done is the answer recorded.
**Manifest input defaults to one Shaped Ticket for the whole Manifest.** Its Deliverables remain an internal execution and verification structure. Split into one Ticket per Deliverable only when the caller explicitly asks for delegation or parallel pickup. Existing Deliverable boundaries are the smallest allowed split: if one is too large, or is a layer rather than something exercisable end to end, amend the Manifest rather than shipping a unit that cannot be judged end to end.
**Question input gets a separate Question Ticket only when the question needs its own lifecycle.** A question that the current work can answer, or that merely records an execution choice, stays inside the containing Ticket. Related questions managed together become one Question Ticket. An explicit request to track or delegate a question establishes that separate lifecycle.
**Follow-up input remains distinct from the source obligation.** Current-scope work stays on the source Ticket; a blocker to its definition of done escalates it. Search the effort's open set first, fold the finding into an existing Ticket when it already covers the work, and group related new findings before authoring. Every emitted follow-up links back to its source and carries the same effort membership unless the finding genuinely belongs elsewhere.
A Ticket is a contract with a stranger who may not have manifest-dev. Rewrite the source into the convention's anatomy rather than excerpting it. Could a competent stranger holding only this Ticket understand why the work exists, know its bounds, avoid its traps, and judge it done? Manifest-dev vocabulary in an emitted Ticket fails that test.
For Manifest input, translate the full coherent outcome into one Ticket. Problem and Goal become Why; Appetite and Out of bounds become Scope; all applicable Global Invariants become plain Rules that must hold; risks and assumptions become Watch out for; the Initial Approach remains optional advice; every Deliverable and gate contributes to one plain-prose Definition of done. Gate kinds, codes, and evaluator instructions stay behind.
In explicit split mode, apply the same translation to each Deliverable and copy every applicable Global Invariant into every Ticket. A Deliverable that is a layer rather than something exercisable end-to-end is amended in the Manifest first, per the cut rule above. Do not emit the Manifest's ceiling invariant as a rule; its substance already lives in each Ticket's Scope.
For an ordinary new Ticket, grant Auto only when neither doing the work nor judging it done needs human knowledge, taste, or authority, and the author chooses to trust unattended execution. When in doubt, withhold.
For a follow-up, first establish whether this `ticket-up` authoring boundary has a fresh human grant for Auto on the new Ticket. A person directly authoring here, or explicitly reviewing and authorizing Auto here, may grant the follow-up independently of its source after the follow-up passes the normal Auto criterion. The person still chooses whether to trust unattended execution; Shaped never implies Auto.
Without that fresh human grant, preserve authority rather than widening it: the source Ticket must carry Auto **and** the follow-up must independently pass the same grant criterion. An unattended or nested authoring step cannot turn an ungranted source into future unattended work. Merely invoking `run-ticket` manually on an ungranted source is not a fresh grant for follow-ups discovered inside that run; a person must separately authorize Auto at this authoring boundary.
N
Skills for agentic coding CLIs. They keep three things in your project instead of in your head: what it's becoming, what's worth doing next, and what done means here. The agent reads them, works against them, and checks the result before reporting it finished.
Harden a manifest-dev task guidance file for one-shot quality — either /define''s quality-gate/Default set or figure-out''s probe set. Iterates: orthogonality…
Analyze Claude Code sessions to learn what went right/wrong and suggest high-confidence improvements to skills. Use when asked to analyze a session, learn from…
Sync the prompt-engineering plugin from a local clone of claude-code-plugins into .claude/ so the repo is self-contained for isolated/web environments. Copies…
Regenerate the Codex distribution copy of the plugin skills and check it for drift. Run after changing anything under claude-plugins/*/skills. OpenCode and Pi…
Analyzes the current project structure and tech stack. Use when asked to explore, understand, or summarize a project. Trigger terms: project overview, analyze…
Author-side PR lifecycle babysitter and companion to review-pr. Use when the user wants to tend an existing GitHub PR through CI, review threads, description…