agentleFS
kambleakash0kambleakash0/agent-skills/skills/slice-the-spec/SKILL.md

slice-the-spec

Skill10 starsChanged 7 months agoMITView on GitHub

Turn a PRD into a Kanban-ready backlog of vertically sliced issues, with clear dependencies and HITL/AFK flags.

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user:

    line 18
  2. Very specific:

    If there is no PRD yet or the requirements are fuzzy, suggest using /grill-me and /spec-writer first.

    line 24
  3. Distinctive:

    Prefer AFK slices where possible, but mark HITL clearly where human judgment is essential.

    line 32
  4. Distinctive:

    If multiple PRDs are present, confirm with the user which one to use.

    line 41
  5. Distinctive:

    Avoid horizontal slices like “set up database tables” or “build all UI components” unless truly necessary.

    line 49
  6. Distinctive:

    Should any slices be merged or split?

    line 75
  7. Distinctive:

    Every issue should be implementable by someone who has the PRD and the repo, with minimal additional clarification.

    line 92
  8. Distinctive:

    Prefer vertical slices that “light up” a narrow but real path through the system over broad foundational work that shows no user‑visible …

    line 93
  9. Distinctive:

    Always expose and annotate HITL work.

    line 94
  10. Very specific:

    If the PRD is missing critical details, ask a small number of targeted clarification questions instead of guessing, or suggest returning to …

    line 95

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Turn a PRD into Issues

This skill takes a clarified PRD and turns it into a small, sharp backlog of vertically sliced GitHub issues. Each issue should be an independently grabbable “tracer bullet” that cuts through all relevant layers end‑to‑end, flushing out unknown unknowns as early as possible.

When to Use

Use this skill when the user:

  • Has a PRD or similar spec (possibly created via /spec-writer) and now wants concrete issues.
  • Wants to turn “destination” docs into a Kanban board of tasks.
  • Needs to break work down so multiple engineers or agents can work in parallel without stepping on each other.

If there is no PRD yet or the requirements are fuzzy, suggest using /grill-me and /spec-writer first.

Concepts

  • Vertical slice / tracer bullet: A thin, end‑to‑end piece of functionality that goes through all necessary layers (data, domain, API, UI, etc.), not just one layer in isolation.
  • HITL issue: Requires Human‑In‑The‑Loop decisions (design review, stakeholder sign‑off, risky refactor, etc.).
  • AFK issue: “Away‑From‑Keyboard” for the human; safe enough for an autonomous agent to implement and merge without human intervention.

Prefer AFK slices where possible, but mark HITL clearly where human judgment is essential.

Workflow

You may compress or skip steps if the context already provides the answer (e.g., the PRD is already loaded and clearly scoped).

  1. Locate and understand the PRD
    • Find the PRD from the conversation or repository (issue, doc, file).
    • Skim to understand: goals, user stories, functional requirements, and explicit non‑goals.
    • If multiple PRDs are present, confirm with the user which one to use.
  2. Explore the codebase
    • Inspect relevant parts of the repo to understand current architecture, major modules, and likely integration points.
    • Note any existing patterns you should respect (e.g., existing feature flags, background jobs, notification systems).
  3. Draft vertical slices (tracer bullets)
    • Break the PRD into a small set of vertical slices, each of which delivers end‑to‑end value and exercises multiple layers of the system.
    • Avoid horizontal slices like “set up database tables” or “build all UI components” unless truly necessary; fold those into end‑to‑end slices instead.
    • For each slice, decide whether it is HITL or AFK, preferring AFK when risk is low.
  4. Define issue structure and dependencies For each slice, define:

    • Title: Short, descriptive, directly tied to user value.
    • Type: HITL or AFK.
    • Description: What this issue will deliver, referencing specific user stories and PRD sections.
    • Acceptance criteria: Clear, testable criteria including edge cases.
    • Blocked by: Which other slices must be completed first (if any).

    Identify at least one unblocked issue that can be started immediately.

  5. Quiz the user and iterate Present the proposed breakdown as a numbered list of slices with:

    • Title
    • Type (HITL/AFK)
    • Blocked by
    • User stories covered

    Ask the user:

    • Does the granularity feel right (too coarse, too fine)?
    • Are dependencies correct?
    • Should any slices be merged or split?
    • Are the right slices marked as HITL vs AFK?

    Refine until the user approves the breakdown.

  6. Generate GitHub‑ready issues
    • For each approved slice, generate a GitHub‑ready issue body (Markdown) including:
      • Context summary
      • Detailed task description
      • Acceptance criteria
      • Links to the PRD and related issues
      • HITL/AFK flag and any notes for human reviewers
    • Optionally group them into epics or labels if the repo conventions are clear.

Behavior and Rules

  1. Optimize for few, high‑leverage slices, not dozens of micro‑issues.
  2. Every issue should be implementable by someone who has the PRD and the repo, with minimal additional clarification.
  3. Prefer vertical slices that “light up” a narrow but real path through the system over broad foundational work that shows no user‑visible change.
  4. Always expose and annotate HITL work; do not hide human decisions inside AFK issues.
  5. If the PRD is missing critical details, ask a small number of targeted clarification questions instead of guessing, or suggest returning to /spec-writer.

Example

Input: A PRD describing a new weekly summary email feature for team admins, with detailed user stories and requirements.

You:

  1. Locate the PRD file or issue, skim goals, user stories, and non‑goals.
  2. Explore the repo to find existing notification and email infrastructure.
  3. Propose 3–5 vertical slices such as:
    • “Engine for computing weekly admin metrics (AFK)”
    • “Admin‑configurable summary email template (HITL)”
    • “Schedule and send weekly summary emails (AFK)”
    • “Admin UI to preview and enable weekly summaries (HITL)”
  4. Present these with types, dependencies, and mapped user stories; refine based on the user’s feedback.
  5. Emit final GitHub issue bodies for each slice, ready to paste or create via API.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.