Skip to main content

to-issues

Use when the user wants to turn a plan, spec, PRD, or approved idea into independently grabbable issue-tracker slices. Not when the work still needs a PRD, repository discovery, or a generic handoff artifact.

Version 1.0.0 draft GNU GPL v3

Last synced:

Version
1.0.0
Maturity
draft
Repository
agent-skills
License
GNU GPL v3

Skill metadata

Repository
matt-riley/agent-skills
Source file
skills/to-issues/SKILL.md
Version
1.0.0
Maturity
draft
Compatibility
Agent Skills-compatible coding agents.
License
GNU GPL v3

SKILL.md

To Issues

Use this skill when the goal is to break already-shaped work into issue-tracker tickets that can be picked up independently. Favor tracer-bullet vertical slices: thin end-to-end slices that are demoable or verifiable on their own, with explicit dependencies and a clear distinction between AFK work and HITL work.

Use this skill when

  • The user wants to turn a plan, spec, PRD, or approved feature idea into implementation issues.
  • The next step is creating dependency-aware tickets rather than refining the product narrative.
  • The work should be split into small vertical slices instead of broad layer-by-layer tasks.
  • The output should either be published to the issue tracker or left as ready-to-paste issue bodies.

Do not use this skill when

  • The work still needs a PRD or broader product framing before it can be split into tickets — use to-prd.
  • The codebase is not understood well enough to describe the slices responsibly — use acquire-codebase-knowledge.
  • The user needs a reusable markdown handoff artifact rather than tracker-native issues — use workflow-contracts.
  • The ask is still too vague to tell whether issue slicing is the right next step — use reverse-prompt.

Routing boundary

Situation Use this skill? Route instead
Break an existing plan or PRD into implementation tickets with dependencies Yes -
Create the product-facing requirements document before any ticket breakdown exists No to-prd
Produce a structured markdown handoff for planning, review, or execution No workflow-contracts
Explore the repository because the current architecture is still unclear No acquire-codebase-knowledge
Rewrite an under-specified ask before deciding whether tickets should exist yet No reverse-prompt

Inputs to gather

Required before editing

  • The source material to break down: plan, spec, PRD, approved issue, or a clearly scoped feature brief.
  • Whether the result should be published now or left as a reviewed draft.
  • The target issue tracker context, including any known parent issue or epic link.

Helpful if present

  • User stories or acceptance criteria already captured in the source material.
  • Existing domain vocabulary, ADRs, or repository conventions for the affected area.
  • Known issue template, labels, or triage conventions for the target project.

Only investigate if encountered

  • Missing architectural context that changes where slice boundaries should fall.
  • Existing issue comments, linked tickets, or parent issue notes that materially change dependencies.
  • Missing permissions, labels, or tracker configuration needed for publication.

First move

  1. Read the source material and current conversation before asking for anything new.
  2. Pull just enough repository context to make the slice boundaries and terminology credible.
  3. Draft the first vertical-slice breakdown before asking the user to adjust granularity or dependencies.

Workflow

  1. Start from the source artifact already in context. If the user supplied an issue reference, fetch its body and comments before slicing.
  2. Identify the outcomes, user stories, and scope boundaries that the issues need to cover.
  3. Break the work into tracer-bullet vertical slices using references/slicing-guide.md:
    • each slice should cut through the necessary layers end-to-end
    • each slice should be demoable or verifiable on its own
    • prefer many thin slices over a few thick ones
  4. For each slice, draft:
    • title
    • type: AFK or HITL
    • dependencies / blockers
    • user stories covered
    • a concise "What to build" summary
    • concrete acceptance criteria
  5. Present the proposed breakdown as a numbered list and get explicit approval on:
    • granularity
    • dependencies
    • merge or split suggestions
    • AFK vs HITL classification
  6. Iterate until the slice breakdown is approved.
  7. If publication is requested, create issues in dependency order using assets/issue-template.md so blocker references point to real issue IDs. Apply only verified labels or tracker fields.

Outputs

  • An approved vertical-slice issue breakdown.
  • One ready-to-paste issue body per slice using assets/issue-template.md.
  • Either published issues in dependency order or a complete draft set plus any explicit publication blocker.

Guardrails

  • Keep slices vertical, not horizontal. Do not create separate tickets like "database work", "API work", and "UI work" when one thin end-to-end slice would be more executable.
  • Prefer AFK over HITL unless human review, design, or decision-making is genuinely required.
  • Do not publish issues before the user approves the proposed breakdown.
  • Do not invent labels such as needs-triage, project fields, or tracker conventions unless they are verified in the target project.
  • Do not close or modify a parent issue as part of this skill.
  • If the source material is still too fuzzy to slice responsibly, stop and route to to-prd or reverse-prompt instead of guessing.

Validation

  • Run node skills/skill-authoring/scripts/validate-skill-library.mjs skills/to-issues/SKILL.md.
  • Confirm every slice has a title, AFK/HITL type, dependency status, user-story coverage, and acceptance criteria.
  • Confirm the proposed issues are thin vertical slices rather than layer-based tasks; use references/checklist.md.
  • Confirm the publish order matches dependency order and does not create circular blockers.
  • Confirm any tracker labels, templates, or metadata are applied only when they are verified.
  • should trigger: "Break this approved PRD into implementation issues and open them in dependency order."
  • should not trigger: "Write the PRD for this feature before we decide how to split the work."

Examples

  • "Turn this implementation plan into independently grabbable GitHub issues, and show me the breakdown before you publish anything."
  • "Break this PRD into thin vertical slices with blocker relationships and mark which ones need human review."
  • "Use the current issue and repo context to draft issue bodies for the next implementation wave, but leave them as ready-to-paste markdown."

Reference files