Skip to content
Back to skills

Plan To Tickets

ASecurity

Use when a large plan, PRD, feature, refactor, research plan, or multi-step coding task must be split into small ready-for-agent tickets with acceptance criteria, verification commands, blockers, and vertical tracer-bullet slices. Do not use for small tasks that should be implemented directly, single-bug fixes, or chat-only summaries.

  • 154 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 8, 2026
ai-agentspythongoshellrefactoringbackend

Security analysis

A100/100

Scanned October 8, 2026

npx -y skills add AnastasiyaW/codex-claude-code-config --skill plan-to-tickets --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plan To Tickets?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Plan To Tickets
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/anastasiyaw-plan-to-tickets/badge)](https://www.skillsdirectory.com/skills/anastasiyaw-plan-to-tickets)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: plan-to-tickets
description: Use when a large plan, PRD, feature, refactor, research plan, or multi-step coding task must be split into small ready-for-agent tickets with acceptance criteria, verification commands, blockers, and vertical tracer-bullet slices. Do not use for small tasks that should be implemented directly, single-bug fixes, or chat-only summaries.
---

# Plan To Tickets

Turn a large plan into small, independently executable tickets. Use this before handing a broad plan to agents, before long autonomous work, or when the user asks for tickets/issues/slices.

## Output Location

If no issue tracker is explicitly selected, write local tickets under:

`<project>/.agent/tickets/<YYYY-MM-DD>-<slug>/`

Each ticket is one Markdown file:

`TICKET-001-short-slug.md`

## Required Ticket Shape

Each ticket must include these headings exactly:

- `## Status`
- `## Parent`
- `## What To Build`
- `## Acceptance Criteria`
- `## Verification`
- `## Blocked By`
- `## Notes`

`## Status` must contain `ready-for-agent` only when all acceptance criteria and verification steps are concrete.

`## Acceptance Criteria` must contain checklist items (`- [ ] ...`).

`## Verification` must contain at least one command, script, test, artifact check, or explicit manual-review gate.

## Slicing Rules

- Prefer vertical tracer bullets: a narrow complete path through the required layers, not "all backend" then "all UI" then "all tests".
- A ticket must be independently verifiable.
- Put prefactoring first only when it makes a later slice smaller or safer.
- Keep dependencies explicit in `## Blocked By`; write `None` when it can start immediately.
- Do not include stale-prone file paths unless the path is already a stable contract or current codebase evidence.
- Do not publish external tracker issues unless the user asked for that tracker.

## Workflow

1. Read the plan/spec and current repo context needed to avoid fake tickets.
2. Draft ticket files in dependency order.
3. Run the validator:

```powershell
python "$env:USERPROFILE\.claude\claude-code-config\scripts\validate_agent_tickets.py" --tickets-dir <ticket-dir>
```

4. Fix every validator failure before claiming the ticket split is ready.
5. Report the ticket directory and the count of ready tickets.

## Gotchas

- Horizontal tickets overload later agents. If a ticket cannot be verified alone, split it differently.
- "Implement feature" is not an acceptance criterion. ACs must describe observable behavior or artifacts.
- "Run tests" alone is weak verification. Name the test command or the artifact that proves success.
- Do not use this to postpone work. Ticketization is for large plans; small tasks should be implemented directly.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.