wf-shape
This skill should only be used when the user uses the word workflow and asks to shape a project from messy inputs into a de-risked, de-scoped shaped packet (brief, breadboard, risks, spikes) ready for wf-plan, with handoff/pickup boundaries to avoid context rot.
Best use case
wf-shape is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
This skill should only be used when the user uses the word workflow and asks to shape a project from messy inputs into a de-risked, de-scoped shaped packet (brief, breadboard, risks, spikes) ready for wf-plan, with handoff/pickup boundaries to avoid context rot.
Teams using wf-shape should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/wf-shape/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How wf-shape Compares
| Feature / Agent | wf-shape | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
This skill should only be used when the user uses the word workflow and asks to shape a project from messy inputs into a de-risked, de-scoped shaped packet (brief, breadboard, risks, spikes) ready for wf-plan, with handoff/pickup boundaries to avoid context rot.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# wf-shape ## Purpose Turn messy inputs into a shaped packet that is ready for commit-ready planning (wf-plan) or, for small/low-risk work, ready to go straight into wf-develop driven by `prd.json`. Optimise for: - picking the problems/opportunities that create high user and/or business value - de-scope (draw the perimeter) - risks understood (rabbit holes surfaced + treated) - concrete artefacts (brief + breadboard + spikes) - PRD spine that agents can actually execute against ## Inputs - Raw notes, feature idea, bug report, links - Constraints + appetite/timebox - Any existing artefacts (designs, metrics, incidents) ## Outputs (dossier) Create/update a dossier folder: - `docs/04-projects/<lane>/<id>_<slug>/` Files: - `brief.md` (1–2 pager) - `breadboard-pack.md` - `risk-register.md` - `spike-investigation.md` (optional; single file) - `prd.md` (can be one or many) - `prd.json` (recommended; created via `create-json-prd`) (can be one or many) ## Steps 0) Pickup (recommended if resuming / new thread) - Understand where we are in the process. - If new, start at step 1. - Otherwise start at the relevant step (usually step 8), but first ensure previous steps have been completed. 1) Start a dossier (if not created already) - Choose `id_<slug>` (prefer `0001_<short>`). - Choose a lane under `docs/04-projects/`. - Create dossier + stub files. 2) Write the 1–2 pager brief - Inputs should come from the user by invoking `ask-questions-if-underspecified` and/or using relevant context. - Write `brief.md`. - Hard requirements: - goals + non-goals - in-scope/out-of-scope perimeter - top risks + unknowns - open questions - GO/NO-GO placeholder 3) Breadboard - Invoke breadboarding skill. - Save as `breadboard-pack.md` (places/affordances/connections + parts list + rabbit holes + fit check). 4) Risk register - Write `risk-register.md`. - For every top risk choose: Cut / Patch / Spike / Out-of-bounds. 5) Spike investigation (only for Spike items) - Invoke spike-investigation skill for each Spike. - Record results in `spike-investigation.md`. 6) Oracle bundle (mandatory) - Invoke `oracle` once to create an oracle bundle for the dossier so later oracle passes have clean, repeatable context. - Bundle inputs should include (at minimum): - dossier `brief.md` - dossier `breadboard-pack.md` - dossier `spike-investigation.md` - dossier `risk-register.md` - any raw notes/links that materially affect scope - any existing artefacts (designs, metrics, incidents) that materially affect decisions 7) Midpoint handoff (mandatory, before PRD creation) - Invoke `handoff` skill as a context checkpoint (this is separate from the final handoff). - Include: - dossier path - current brief status + perimeter draft (in/out) - all context from dossier files - oracle bundle created: y/n + what was included - next step: PRD creation 8) Pickup on context - User manually links to the midpoint handoff file and oracle findings and invokes the `pickup` skill. 9) Review oracle pass and make updates to: - `brief.md` - `breadboard-pack.md` - `spike-investigation.md` (when present) - `risk-register.md` Also consider making updates to relevant `docs/03-architecture` documents where required. 10) PRD (source of truth for “done”) - Use the `create-prd` skill to create one or many `prd.md` in the dossier. - If acceptance criteria are missing, write explicit TODOs rather than vibes. 11) JSON PRD (recommended) - Run `create-json-prd` skill so agents/tools can execute acceptance criteria. - Each `prd.md` should have a corresponding `prd.json`. - Ensure it validates against the repo schema (NO-GO if invalid). 12) Synthesis + perimeter lock - Update `brief.md`, `prd.md`s and `prd.json`s to reflect the final perimeter, cuts, and risk treatments. - Ensure breadboard is buildable as parts. 13) Shaping decision - GO only if biggest risks are proved, cut, or out-of-bounds. 14) Handoff (pick wf-plan vs wf-develop explicitly) - Invoke `handoff` and include: - dossier path - perimeter (in/out) - biggest risks + spike outcomes - PRD status (`prd.json` validated? y/n) - **Plan needed: yes/no + why** - Recommended paths: - If Plan needed = yes → wf-plan - If Plan needed = no → wf-develop using `prd.json` - Recommend starting the next step in a fresh thread: - `/new` then `pickup` then run wf-plan or wf-develop with the dossier path ## Verification - brief has perimeter + testable outcomes - risk register has a mitigation for each top risk - PRD exists with acceptance criteria - `prd.json` validates (when created) ## Go/No-Go - GO if a stranger can explain what will be built and what won’t from brief + PRD + breadboard. - NO-GO if core mechanics are still foggy.
Related Skills
skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
modular-skills-architect
Map and refactor an agent context ecosystem: skills, commands/workflows, hooks, agent files, AGENTS.md templates, and docs. Output system map, module/dependency design, Register updates, and a concrete split/consolidate/rename/delete plan. Use when routing or ownership is messy.
heal-skill
This skill should be used when fixing incorrect SKILL.md files with outdated instructions or APIs.
create-agent-skills
Expert guidance for creating, writing, and refining Claude Code Skills. Use when working with SKILL.md files, authoring new skills, improving existing skills, or understanding skill structure and best practices.
agent-native-audit
Comprehensive agent-native architecture audit with scored principles and multi-slice review. Use for system-wide health checks or periodic audits.
write-judge-prompt
Design LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness). Do NOT use when the failure mode can be checked with code (regex, schema validation, execution tests). Do NOT use when you need to validate or calibrate the judge — use validate-evaluator instead.
validate-evaluator
Calibrate an LLM judge against human labels using data splits, TPR/TNR, and bias correction. Use after writing a judge prompt (write-judge-prompt) when you need to verify alignment before trusting its outputs. Do NOT use for code-based evaluators (those are deterministic; test with standard unit tests).
generate-synthetic-data
Create diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation. Use when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. Do NOT use when you already have 100+ representative real traces (use stratified sampling instead), or when the task is collecting production logs.
evaluate-rag
Guides evaluation of RAG pipeline retrieval and generation quality. Use when evaluating a retrieval-augmented generation system, measuring retrieval quality, assessing generation faithfulness or relevance, generating synthetic QA pairs for retrieval testing, or optimizing chunking strategies.
eval-audit
Audit an LLM eval pipeline and surface problems: missing error analysis, unvalidated judges, vanity metrics, etc. Use when inheriting an eval system, when unsure whether evals are trustworthy, or as a starting point when no eval infrastructure exists. Do NOT use when the goal is to build a new evaluator from scratch (use error-analysis, write-judge-prompt, or validate-evaluator instead).
error-analysis
Help the user systematically identify and categorize failure modes in an LLM pipeline by reading traces. Use when starting a new eval project, after significant pipeline changes (new features, model switches, prompt rewrites), when production metrics drop, or after incidents.
build-review-interface
Build a custom browser-based annotation interface tailored to your data for reviewing LLM traces and collecting structured feedback. Use when you need to build an annotation tool, review traces, or collect human labels.