Skip to content
Back to skills

Prd

ASecurity

Convert current conversation context into a structured Product Requirements Document (PRD) and optionally post it to GitHub Issues. Use before /supergraph:plan when requirements came from a conversation, user story, or feature discussion rather than a formal spec.

  • 22 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added August 31, 2026
ai-agentsgobashtestinggitsecurity

Security analysis

A100/100

Scanned August 31, 2026

npx -y skills add datit309/supergraph --skill prd --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Prd?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Prd
[![Security: A β€” Skills Directory](https://www.skillsdirectory.com/api/skills/datit309-prd/badge)](https://www.skillsdirectory.com/skills/datit309-prd)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: prd
description: Convert current conversation context into a structured Product Requirements Document (PRD) and optionally post it to GitHub Issues. Use before /supergraph:plan when requirements came from a conversation, user story, or feature discussion rather than a formal spec.
---

# /supergraph:prd

Turn conversation into a machine-readable PRD. Gives `/supergraph:plan` a concrete spec to work from.

Announce: "πŸ“‹ /supergraph:prd β€” extracting requirements..."

## Steps

### 1. Read CONTEXT.md (if exists)

```bash
head -60 CONTEXT.md 2>/dev/null || echo "No CONTEXT.md"
```

Use domain vocabulary from CONTEXT.md in the PRD β€” never invent new terms for existing concepts.

### 2. Extract requirements from conversation

Gather from current conversation context:
- What the user wants to accomplish
- Any constraints or non-goals mentioned
- Technical decisions already made
- Edge cases or concerns raised

Ask one clarifying question if a critical piece is missing. Do not ask multiple questions at once.

### 3. Write PRD

```markdown
# PRD: [Feature/Fix Name]

## Problem
[One paragraph: what is broken or missing, and why it matters]

## Solution
[One paragraph: the approach β€” what will be built]

## User Stories
- As a [role], I want [action] so that [outcome]
- (2-5 stories max)

## Acceptance Criteria
- [ ] [Observable, testable condition]
- [ ] [Observable, testable condition]
- (Each maps to a test in /supergraph:plan)

## Implementation Decisions
- [Technology/pattern chosen and why]
- [What was explicitly ruled out and why]

## Testing Decisions
- [Unit: what gets unit tested]
- [Integration: what needs integration tests]
- [E2E: what needs Playwright coverage, if any]

## Out of Scope
- [Explicitly not included in this work]

## Notes
- [Domain terms: add to CONTEXT.md if new]
- [Dependencies on other work]
- [Security considerations]
```

### 4. Present for approval

Show PRD to user. Ask: "Does this capture what you need? [yes / adjust]"

Incorporate feedback, re-present if needed.

### 5. Post to GitHub Issues (optional)

If user confirms and repo has GitHub remote:
```bash
gh issue create \
  --title "[PRD] <feature name>" \
  --body "$(cat <<'EOF'
[PRD content]
EOF
)" \
  --label "enhancement,ready-for-agent"
```

Report issue URL.

### 6. Save locally

Save to `docs/supergraph/plans/YYYY-MM-DD-prd-<slug>.md`.

### 7. Update CONTEXT.md

If the PRD introduced new domain terms:
```bash
# Append to CONTEXT.md
printf '\n## <term>\n[definition from PRD]\n' >> CONTEXT.md
```

### 8. Report

```
βœ… /supergraph:prd complete
- PRD: docs/supergraph/plans/YYYY-MM-DD-prd-<slug>.md
- Issue: #N (if posted)
- Acceptance criteria: N
- Next: /supergraph:plan
```

## Rules

- Never skip user approval (step 4)
- Acceptance criteria must be observable and testable β€” no vague criteria
- Out of scope section is mandatory β€” prevents scope creep in plan
- Post to GitHub only if user has confirmed the PRD

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.