agentleFS
Beakpoint-InsightsBeakpoint-Insights/roost/skills/shape-epic/SKILL.md

shape-epic

Skill32 starsChanged 17 days agoMITView on GitHub

Shape a product idea into a fully-defined, buildable epic. Composable phases from problem statement through epic definition, design (a UI prototype, an architecture PR, or both), slice planning, test creation, build guide, and a tech lead handoff prompt attached to the epic. Can enter at any phase depending on what already exists. Use when user wants to go from idea to implementation-ready epic.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When invoked without a specific phase, inspect what exists and recommend where to start:

    line 56
  2. Distinctive:

    No epic key provided: Ask if they have an existing epic or want to start from a problem statement.

    line 58
  3. Distinctive:

    When entering mid-stream, always inventory what exists before proposing new work.

    line 96
  4. Distinctive:

    Before declaring a phase complete:

    line 102

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Shape Epic

Turn a product idea into a fully-defined, buildable epic with its design (a UI prototype, an architecture or design-doc PR, or both), ordered slices, test scenarios, a visual build guide, and a tech lead prompt that hands the epic to developer agents.

Trigger

  • /shape-epic — auto-detect what exists, resume from there
  • /shape-epic PROJ-XXXX — shape an existing epic
  • /shape-epic problem — start from problem analysis
  • /shape-epic design (or prototype) — design an existing epic: a UI prototype, an architecture PR, or both
  • /shape-epic slices — plan slices for an epic that has its design
  • /shape-epic tests — create test scenarios for existing slices
  • /shape-epic guide — generate the build guide from existing slices and tests
  • /shape-epic handoff — write the tech lead prompt and attach it to the epic

Team config

First, load the team config with roost-config. The tracker and its project are in tracker (its adapter says how keys look), the design system is design, the repos the tech lead reads are repos, and the roadmap question uses roadmap.

Rules for every phase

  • End every stop with the exact next step, as a question. When you finish a phase or wait for feedback, the last line names what happens next by phase and skill, so the user knows what "yes" starts: "Next is Phase 4 (slices): I'll split PROJ-2961 into stories from the prototype. Start it, or change the prototype first?" Never end on "let me know" or "what next?".
  • Every question comes with your recommendation and why. List the options with the recommended one first, so the user can answer with a letter. A question without a recommendation costs them a round trip.
  • Scope changes only when the user decides. Anything you find while shaping that isn't part of the epic (a bug in existing code, a refactor, a gap in another feature) goes to the user as a proposed ticket: what it is, the evidence (file:line), and a recommendation (its own Bug, under this epic, or leave it). Don't fold a fix into the design, the slices or the description until they decide, and file the ticket once they do.
  • Work no repo session can do (a PR to another company's repo, a video, a partner conversation, a launch post) is a person-owned task: see phase 4.

Phases

The skill has seven composable phases. Each phase produces artifacts that feed the next. You can enter at any phase if the prerequisites exist.

problem → epic → design → slices → tests → guide → handoff
PhaseProducesPrerequisitesSteps
1. ProblemStructured problem statementAn idea, complaint, or observationreferences/phase-1-problem.md
2. EpicEpic in the tracker, with a rich descriptionProblem statement (or user provides context)references/phase-2-epic.md
3. DesignThe target state: an interactive HTML prototype for UI work, an architecture or design-doc PR for how the system is built or run, or bothEpic description with goals and scopereferences/phase-3-design.md
4. SlicesOrdered stories under the epicEpic + designreferences/phase-4-slices.md
5. TestsThe tests for every slice, in the form the team's tests takeSlices (+ design for testable behaviors)references/phase-5-tests.md
6. GuideVisual build-guide artifactSlices + testsreferences/phase-6-guide.md
7. HandoffTech lead prompt, attached to the epic with the guide and the prototype (if any)Slices + tests + guide + designreferences/phase-7-handoff.md

Before starting a phase, read its steps file in full and follow it. Each file holds that phase's process, guidance and expected output.

Auto-Detection

When invoked without a specific phase, inspect what exists and recommend where to start:

  1. No epic key provided: Ask if they have an existing epic or want to start from a problem statement.
  2. Epic key provided: Fetch the epic and its children from the tracker.
    • Has stories under it? Skip to phase 5 or 6.
    • Has tests? Skip to phase 6.
    • Has a build guide but no <EPIC> Tech Lead Prompt.md attachment? Skip to phase 7.
    • Has only a description? Start at phase 3 (design) or phase 4 (slices).
  3. Present what you found and propose the next phase. Let the user override.
I found PROJ-1768 with:
  ✓ Rich epic description (goals, scope, technical context)
  ✓ 5 stories (PROJ-1771 through 1775)
  ✓ 9 test issues
  ✗ No build guide artifact

Recommended next step: Phase 6 (Build Guide)
Want to proceed, or would you rather revisit an earlier phase?


Resuming Partially-Defined Epics

This skill is designed for epics at any stage of definition. Common entry points:

SituationStart At
"I have an idea but nothing in the tracker"Phase 1 (Problem)
"I have an epic but it's just a title"Phase 1 (Problem), then update the epic in Phase 2
"Epic is well-defined but has no design"Phase 3 (Design)
"We have stories but no tests"Phase 5 (Tests)
"Everything exists, need the handoff doc"Phase 6 (Guide)
"Guide exists, developers need to start"Phase 7 (Handoff)
"We need to add more slices to an existing epic"Phase 4 (Slices), preserving existing ones
"Tests exist but don't match the current slices"Phase 5 (Tests), reviewing what exists first

When entering mid-stream, always inventory what exists before proposing new work. Don't recreate what's already there.


Quality Checklist

Before declaring a phase complete:

  • User has reviewed and approved the output
  • All tickets have rich descriptions (not just titles)
  • Stories are ordered and independently shippable
  • Tests cover every acceptance criterion in every slice
  • Cross-cutting tests are identified and mapped to their last dependency
  • The build guide accurately reflects the current state of all slices and tests
  • The epic has the build guide, the prototype (if it has one) and the tech lead prompt attached, and a comment linking any design PR
  • The tech lead prompt has no unfilled placeholders, and every "In this epic" line comes from this epic's own sources

References

  • See references/epic-description-template.md for the recommended epic description structure
  • See references/slice-checklist.md for the slice quality checklist
  • See references/tech-lead-prompt-template.md for the Phase 7 prompt template and how to fill it
  • See references/phase-<n>-<name>.md for each phase's steps
  • The tests adapter's "Create a test" says how Phase 5 creates tests
  • The skill named in roadmap.skill, if any, puts an epic on the roadmap

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.