Shape
Skill VRIL-LABS/skill-jam/skills/core-collections/spellbook-master/spellbook-master/skills/shape
Shape a raw idea into something buildable. Product + technical exploration. Spec, design, critique, plan. Output is a context packet. Use when: "shape this", "write a spec", "design this feature", "plan this", "spec out", "context packet", "technical design". Trigger: /shape, /spec, /plan, /cp.From its SKILL.md
npx -y skills add VRIL-LABS/skill-jam --skill shapeAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.8 KB, 833 tokens by cl100k_base, as published. Nobody here has run it
/shape
Shape a raw idea into something buildable. Output is a context packet — the unit of specification that precedes implementation.
Workflow
Phase 1: Understand
Accept: raw idea, backlog.d/ item, issue ID, or observation.
Spawn parallel sub-agents to gather context fast: one to map the relevant codebase area (files, patterns, constraints), another to search for prior art (how do other projects solve this? check codebase first, then /research). Synthesize their findings before proceeding.
Phase 2: Product Exploration
GATE: Do NOT write code until product direction is locked.
- Investigate — Problem space, user impact, prior art
- Brainstorm — 2-3 approaches with tradeoffs. Recommend one.
- Discuss — One question at a time. Iterate until locked.
- Draft spec — Goal, non-goals, acceptance criteria
Phase 3: Technical Exploration
-
Explore — 3-5 technical approaches. For each: architecture sketch, files to modify, pattern alignment, effort, tradeoffs. Recommend one.
-
Validate — For effort M or larger, spawn the design review bench in parallel: ousterhout reviews for module depth and information hiding, carmack for shippability and over-engineering, grug for complexity. Give each the design summary and ask for a verdict + concerns. If any has blocking concerns, revise the design before proceeding.
-
Discuss — No limit on rounds. Design isn't ready until user says so.
Phase 4: Context Packet
The output of shape. This is what /autopilot and builders consume.
# Context Packet: <title>
## Goal
<1 sentence — what outcome, not mechanism>
## Non-Goals
- <what NOT to do, even if it seems like a good idea>
## Constraints / Invariants
- <things that must remain true before, during, and after>
## Authority Order
tests > type system > code > docs > lore
## Repo Anchors
- `src/auth/middleware.ts` — current pattern to follow
- `tests/auth/` — existing coverage
## Prior Art
- `src/payments/middleware.ts` — similar pattern
## Oracle (Definition of Done)
- [ ] All existing auth tests pass
- [ ] New endpoint returns 200 with valid token
- [ ] Response time < 100ms p99
## Implementation Sequence
1. <first chunk>
2. <second chunk>
## Risk + Rollout
- <how it could fail, how to undo it>
If you can't write an oracle, the goal isn't clear enough. Go back to Phase 2.
Gotchas
- Vague oracles: "It should work" is not an oracle. "These 3 tests pass and this endpoint returns 200" is. See
references/executable-oracles.md. - Checkbox oracles: Prose checklists drift. Write oracles as commands that return pass/fail, not prose that requires interpretation.
- Speccing after building: A context packet written after implementation is documentation, not specification. Spec first.
- 50 repo anchors: If everything is an anchor, nothing is. Pick 3-10 files whose patterns MUST be followed.
- Skipping non-goals: Agents drift toward scope expansion. Non-goals are load-bearing constraints. Write them.
- Over-speccing implementation details: Specify WHAT and WHY. Let the builder figure out HOW. Detailed pseudocode cascades errors.
Principles
- Minimize touch points (fewer files = less risk)
- Design for deletion (easy to remove later)
- Favor existing patterns over novel ones
- YAGNI ruthlessly
- Recommend, don't just list options
- One question at a time
What ships with it: 5 files
87.3 KB alongside SKILL.md
references/
- breadboarding.md59.4 KB
- critique-personas.md1.3 KB
- executable-oracles.md2.3 KB
- shaping-methodology.md22.6 KB
- writing-plans.md1.8 KB