to-tickets

Files (2)

• LICENSE
• SKILL.md

SKILL.md (first lines)

<!--
Modified copy of https://github.com/mattpocock/skills/tree/3cca18b368ae95cdbdebbff572ccafa662551015/skills/engineering/to-tickets
(mattpocock-skills v1.2.3). Copyright (c) 2026 Matt Pocock, MIT — see LICENSE.
Changes: tickets always go to `.agent/tickets/<NN>-<slug>.md` (no tracker path);
numbering is contiguous from 01 in dependency order; acceptance criteria must state
observable behavior; every ticket names its **Seam** (the public interface its
acceptance tests hit) and what it **Touches** (files and functions, for
scheduling) and may carry optional **Risk** and **Suspicions** lines for
the review; trivial slices fold into a neighbour; tickets never carry the
implementation; the approval quiz is skipped when unattended; model invocation is
allowed.
-->

# To Tickets

Break a plan, spec, or conversation into a set of **tickets**: tracer-bullet vertical slices, each declaring the tickets that **block** it.

## Process

### 1. Gather context

Work from whatever is already in the conversation context. If the user passes a reference (a spec path, an issue number or URL) as an argument, fetch it and read its full body and comments.

### 2. Explore the codebase (optional)

If you have not already explored the codebase, do so to understand the current state of the code. Ticket titles and descriptions should use the project's domain glossary vocabulary, and respect ADRs in the area you're touching.

Look for opportunities to prefactor the code to make the implementation easier. "Make the change easy, then make the easy change."

### 3. Draft vertical slices

Break the work into **tracer bullet** tickets.

<vertical-slice-rules>

- Each slice cuts a narrow but COMPLETE path through every layer (schema, API, UI, tests): vertical, NOT a horizontal slice of one layer
- A completed slice is demoable or verifiable on its own
- Each slice is sized to fit in a single fresh context window
- Any prefactoring should be done first
- No trivial tickets: when a slice's whole change is a small guard or fix (a few lines) and no other ticket is blocked by it, fold its acceptance criteria into the neighbouring ticket that touches the same behaviour. Every ticket costs a test-writer, an implementer and a reviewer; a two-line guard doesn't earn three agents

</vertical-slice-rules>

Give each ticket its **blocking edges**: the other tickets that must complete before it can start. A ticket with no blockers can start immediately.

**When in doubt, block.** If you suspect a ticket might need to wait for another (it builds on that ticket's behaviour, touches the same files or functions, or the order just isn't obvious), add the edge. A missing edge costs broken locked tests, rebase churn and a re-run; an extra edge only costs some concurrency.

**Wide refactors are the exception to vertical slicing.** A **wide refactor** is one mechanical change (rename a column, retype a shared symbol) whose **blast radius** fans across the whole codebase, so a single edit breaks thousands of call sites at once and no vertical slice can land green. Don't force it into a tracer bullet; sequence it as **expand–contract**. First expand: add the new form beside the old so nothing breaks. Then migrate the call sites over in batches sized by blast radius (per package, per directory), each batch its own ticket blocked by the expand, keeping CI green batch to batch because the old form still exists. Finally contract: delete the old form once no caller remains, in a ticket blocked by every migrate batch. When even the batches can't stay green alone, keep the sequence but let them share an integration branch that all block a final integrate-and-verify ticket; green is promised only there.

### 4. Quiz the user (attended only)

When a human is in the loop, present the proposed breakdown as a numbered list. For each ticket, show:

- **Title**: short descriptive name
- **Blocked by**: which other tickets (if any) must complete first
- **Seam**: the public interface its acceptance tests hit
- **Touches**: the files and functions it expects to change
- **What it delivers**: the end-to-end behaviour this ticket makes work

Ask the user:

- Does the granularity feel right? (too coarse / too fine)
- Are the blocking edges correct: is any edge missing? (Suspected dependencies are already listed as blockers; say which ones can safely go.)
- Should any tickets be merged or split further?

Iterate until the user approves the breakdown.

Running unattended (no human to answer), skip this step: the approved spec and the plan stage's own reviewers are the checkpoints. Write the tickets directly.

### 5. Write the ticket files

Write one file per ticket under `.agent/tickets/<NN>-<slug>.md`. `NN` is a two-digit number starting at `01`, assigned in dependency order (blockers first) with no gaps and no duplicates: a reader executes them in numeric order and the sequence must never stop early. Delete any stale files in that directory first so the set is exactly the tickets you drafted. Use the template below: one ticket per file, never a single combined file.

`.agent/` is working state, not source. It is never committed; check it is gitignored before committing anything else.

Work the **frontier**: any ticket whose blockers are all done. For a purely linear chain that means top to bottom.

Do NOT close or modify any parent issue.

<ticket-template>

# <NN>: <Ticket title>

**What to build:** the end-to-end behaviour this ticket makes work, from the user's perspective, not a layer-by-layer implementation list.

**Blocked by:** the numbers/titles of the tickets that gate this one, or "None (can start immediately)".

**Seam:** the public interface the acceptance tests hit (a function, an HTTP route, a CLI command, a config loader…), or `none — refactor`.

**Touches:** the files, and the functions within them, this ticket expects to change.

**Risk:** optional. The shared state, persistence, concurrency, money or auth this ticket touches. Omit the line when there is none.

**Suspicions:** optional. What the reviewer should test first: a sibling caller, a key built from the old identity, a guard another ticket added.

- [ ] Acceptance criterion 1
- [ ] Acceptance criterion 2

</ticket-template>

<acceptance-criteria-rule>

Every acceptance criterion states **observable behavior**: something a reviewer can check by running the product, a command, or a test, without reading the implementation. "Running `make test` passes with the new endpoint covered" and "A user with an expired token sees the login page" are criteria. "Add a `TokenExpiry` class" and "Refactor the auth module" are implementation steps and do not belong here.

</acceptance-criteria-rule>

<seam-rule>

Every ticket carries a **Seam** line: the one public interface through which its acceptance criteria are tested, black-box. `implement-spec` hands it to a test-writer that sees nothing else of the code, so name it precisely enough to write a test against, and never an internal helper. A ticket with no behavioral criteria (a pure prefactor or refactor) says `none — refactor`: its existing tests are the contract.

</seam-rule>

<touches-rule>

The **Touches** line is a best guess, for scheduling: `implement-spec` never runs two tickets at once that change the same function, because parallel restructures of one function turn into rebase churn and broken locked tests. Name functions, not only files, when a file is shared (`dispatcher/main.py: _run_pass, _resume_woken`). Two tickets restructuring the same function is also a hint to add a blocking edge between them, or to merge them.

</touches-rule>

<risk-rule>

From jesdi/general-skills, MIT.

More from jesdi/general-skills