Generates Feature Requirements documentation from any combination of inputs -- PRDs, design files/Figma URLs, meeting transcripts, or verbal descriptions. Focuses exclusively on requirements; API contracts and system flows are generated separately after requirements are finalized. Runs in Quick mode (3 contexts, ~15 min) for MVPs/small features or Comprehensive mode (6 contexts, ~30 min) for production/complex features. Use when asked to: generate requirements, write a feature spec, create a PRD, or document a feature.
Generates Feature Requirements documentation from any combination of inputs -- PRDs, design files/Figma URLs, meeting transcripts, or verbal descriptions. Focuses exclusively on requirements; API contracts and system flows are generated separately after requirements are finalized. Runs in Quick mode (3 contexts, ~15 min) for MVPs/small features or Comprehensive mode (6 contexts, ~30 min) for production/complex features. Use when asked to: generate requirements, write a feature spec, create a PRD, or document a feature.
Generate Requirements — Entry Orchestrator
NON-NEGOTIABLE (read first)
Never fabricate data, business rules, or field names. Mark unknowns as [TBD].
Save every output file before presenting in chat. Chat is ephemeral.
Wait for user approval at every checkpoint. Never auto-proceed.
Requirements describe WHAT, never HOW (solution) or WHAT IT LOOKS LIKE (design).
Source traceability on every claim. Unsourced = flag as [TBD].
Critical Rules
Never skip steps — The three workflows must run in order
Always save files — Every output is saved to the workspace, not just shown in chat
Mark TBDs honestly — Never invent business rules, performance targets, or field names
Wait at checkpoints — After each workflow, STOP and wait for user approval before continuing
One skill at a time — Complete each workflow file fully before reading the next
Respect existing patterns — If Swagger is provided, follow its naming and error conventions
🎯 Purpose
This skill generates production-ready Agile requirements documentation from messy inputs.
It orchestrates three sequential workflows, each saved as a file, so context survives across long sessions.
[output-folder] is provided by the user during intake (Step 2). There is no default — always ask.
Note: This skill focuses on Feature Requirements only. API Contracts and System Flows are generated separately after requirements are finalized, using /rest-api-contract-generator and a dedicated system flow skill respectively.
Core principle: Truth over completeness. Mark unknowns as [TBD]. Never fabricate data.
🔗 When Called from /requirements-pipeline (Stage 7)
When this skill is invoked from the /requirements-pipeline, skip Steps 1-3 entirely (workspace scan, project context loading, intake questions, pre-flight checks). The pipeline has already completed all of this.
The pipeline provides:
Feature name -- from Stage 1
Mode -- always Comprehensive
All inputs -- processed and verified in Stages 1-6 (including scenario matrix, assumptions, user flows)
Output folder -- asked by the pipeline at Stage 7 if not already known
Project context -- loaded by the pipeline at Stage 1.1 if project-context.md exists
Current state -- from pipeline Stage 1.5
Go directly to Step 4 (Start Synthesis Workflow) with all values pre-loaded, then read workflows/01-synthesize.md.
Persona Context (uses existing personas; new ones flagged for update)
Compliance Context (uses baseline; overrides noted per feature)
IF project-context.md is NOT found:
Proceed to Step 2 as normal. After the session completes, offer:
💡 This is the first time I've run on this project. Want me to create a
project-context.md file from what I learned? This will save time on
future features.
📋 Step 2: Gather Required Context
IF project-context.md was loaded, ask only what's missing:
I have your project context. For this feature, I need:
1. **Feature name** — What is this feature called?
2. **Feature-specific inputs** — Do you have any of these for this feature?
- [ ] PRD or feature description
- [ ] Design files (Figma URL or images)
- [ ] Swagger changes / new endpoints
- [ ] Meeting transcript or notes
- [ ] Just your description in this chat
3. **New or existing feature?**
- A) New feature (greenfield)
- B) Modifying an existing feature
4. **Scope?**
- A) Quick Mode — lightweight analysis, 3 contexts (Business, Product, UX). Best for MVPs, small changes, internal tools.
- B) Comprehensive Mode — full analysis, 6 contexts (adds Persona, Technical, Compliance). Best for production features, complex integrations, compliance-sensitive work.
- C) Suggest based on inputs
5. **Output folder** — Where should generated documents be saved?
Provide the path to the folder (e.g., `Product Artifacts/Feature Requirements/MyFeature`).
The skill will create `Generated/Internal/` and `Generated/Report/` subdirectories inside it.
IF project-context.md was NOT loaded, ask the full set:
To generate requirements, I need a few details:
1. **Feature name** — What is this feature called?
2. **Inputs available** — Which of these do you have?
- [ ] PRD or feature description document
- [ ] Design files (Figma URL or images)
- [ ] Swagger/OpenAPI specification
- [ ] Meeting transcript or notes
- [ ] Just your description in this chat
3. **New or existing feature?**
- A) New feature (greenfield)
- B) Modifying an existing feature
4. **Scope?**
- A) Quick Mode — lightweight analysis, 3 contexts (Business, Product, UX). Best for MVPs, small changes, internal tools. (~15 min)
- B) Comprehensive Mode — full analysis, 6 contexts (adds Persona, Technical, Compliance). Best for production features, complex integrations, compliance-sensitive work. (~30 min)
- C) Not sure — I'll suggest based on your inputs
5. **Output folder** — Where should generated documents be saved?
Provide the path to the folder (e.g., `Product Artifacts/Feature Requirements/MyFeature`).
The skill will create `Generated/Internal/` and `Generated/Report/` subdirectories inside it.
Create the Generated/Internal/ and Generated/Report/ subdirectories when saving files. The project-context.md stays in the workspace root (not inside the feature folder).
🚀 Step 4: Start Synthesis Workflow
Once intake is complete, confirm what you have and what's pre-loaded:
Project context (if loaded) — mark which contexts are pre-populated vs need feature-specific input
Whether "verify existing" (Step 0.5) should run
Then read the file:
workflows/01-synthesize.md
Follow that file's instructions completely from start to finish.
📎 Related Skills (Complementary)
These existing skills can assist with specific inputs before running this workflow:
Situation
Skill to suggest
User has a meeting recording or .vtt transcript
transcript-to-meeting-notes — process first, use output as PRD input
User has Figma URL or design screenshots
Design analysis is built into this skill's Workflow 1
Requirements are finalized and need API contracts
rest-api-contract-generator — run after requirements are stable
Requirements are finalized and need system flow docs
Future: system-flow-generator — run after requirements are stable
🔄 After All Workflows Complete: Update Project Context
When Workflow 3 (Validation) finishes, offer to create or update the project context:
To capture what was learned in this session (new personas, systems,
constraints, terminology), run /project-context to create or update
your project-context.md.
This is optional but recommended -- it saves time on future sessions.
The /project-context skill handles both first-time creation and incremental updates with source confirmation and conflict detection. Do NOT attempt to create or update project-context.md inline.
INTAKE COMPLETE → Read workflows/01-synthesize.md to begin
