skill-prd

Documentation Community

Repo stars 3,406
Author repo claude-octopus

Interpretation is structured for decision-making; original keeps the upstream SKILL.md unchanged.

Decide Fit First

Core job: Write an AI-optimized PRD using multi-AI orchestration — use when scoping a new feature or product Before writing ANY PRD conten…
Best fit: Use it when the task has reusable inputs, steps, and validation criteria rather than a one-off answer.
Avoid forcing it: If the source lacks commands, platform support, or external-service evidence, keep those fields unknown instead of guessing.

Design Intent

Structure: The skill is organized around “PHASE 0: CLARIFICATION (MANDATORY)”, “PHASE 1: QUICK RESEARCH (Max 2 searches)”, “PHASE 2: WRITE PRD”, “PHASE 2.5: ADVERSARIAL PRD REVIEW (RECOMMENDED)”, showing how the author expects the agent to judge fit, collect context, and produce verifiable output.
Trigger evidence: Prioritize the author’s wording around when to use it, what context to collect, and what output shape to produce.
Evidence boundary: Author text states facts, repository files prove commands and paths, and Fluxly only adds fit, limits, and usage judgment.

How To Use It

Inputs: Provide target material, scope, expected result, forbidden changes, and validation method.
Invocation: Name skill-prd directly; if the source includes slash commands, start with the command and then add task context.
Validation: Start small and check whether the result follows “PHASE 0: CLARIFICATION (MANDATORY) / PHASE 1: QUICK RESEARCH (Max 2 searches) / PHASE 2: WRITE PRD” before expanding.

Boundaries And Review

Dependencies: It usually needs no extra API key, so start with a small validation task.
Permissions: Declared permissions include read / write / shell-exec; ask the agent to state file, command, and rollback boundaries before acting.
Quality bar: A useful result names the deliverable, evidence, and next action. Generic prose means the task needs tighter context.

Host: Codex CLI — This skill was designed for Claude Code and adapted for Codex. Cross-reference commands use installed skill names in Codex rather than /octo:* slash commands. Use the active Codex shell and subagent tools. Do not claim a provider, model, or host subagent is available until the current session exposes it. For host tool equivalents, see skills/blocks/codex-host-adapter.md.

STOP - SKILL ALREADY LOADED

DO NOT call Skill() again. DO NOT load any more skills. Execute directly.

PHASE 0: CLARIFICATION (MANDATORY)

Before writing ANY PRD content, you MUST ask the user these questions:

I need to understand your requirements before creating the PRD.

1. **Target Users**: Who will use this? (developers, end-users, admins, etc.)
2. **Core Problem**: What specific pain point does this solve? Any metrics?
3. **Success Criteria**: How will you measure if this succeeds?
4. **Constraints**: Any technical, budget, or timeline constraints?
5. **Existing Context**: Is this greenfield or integrating with existing systems?

Please answer these (even briefly) so I can create a more targeted PRD.

WAIT for user response before proceeding to Phase 1.

If user says "skip" or provides the feature description inline, extract what you can and note assumptions.

PHASE 1: QUICK RESEARCH (Max 2 searches)

Only search if topic is unfamiliar. Limit to 2 web searches max:

One for domain/market context
One for technical patterns (if needed)

Do NOT over-research. 60 seconds max for this phase.

PHASE 2: WRITE PRD

Structure:

Executive Summary - Vision + key value prop
Problem Statement - Quantified pain points by user segment
Goals & Metrics - SMART goals, P0/P1/P2 priority, success metrics table
Non-Goals - Explicit boundaries (what we WON'T do)
User Personas - 2-3 specific personas with use cases
Functional Requirements - FR-001 format with acceptance criteria
Implementation Phases - Dependency-ordered, time-boxed
Risks & Mitigations - Top 3-5 risks with mitigation strategies

PHASE 2.5: ADVERSARIAL PRD REVIEW (RECOMMENDED)

After drafting the PRD but BEFORE self-scoring, dispatch the draft to a second provider for adversarial review. A single-model PRD has blind spots — cross-provider challenge surfaces wrong assumptions, uncovered scenarios, and contradictory requirements.

Dispatch the PRD draft to a different provider (Codex, Gemini, or Sonnet as fallback) with this prompt:

"Challenge this PRD. What assumptions are wrong? What user scenarios are missing? What requirements contradict each other? What will the first user complaint be? What risk does this PRD ignore?"

After receiving the challenge:

Revise the PRD to address valid challenges
Note dismissed challenges in the Risks section if they have partial merit
Add to PRD footer: Adversarial review: applied

Skip with --fast or when user requests speed over thoroughness. See prd.md command for full dispatch syntax.

PHASE 3: SELF-SCORE

Score against 100-point framework:

AI-Specific Optimization: 25 pts (sequential phases, non-goals, structured format)
Traditional PRD Core: 25 pts (problem statement, goals, personas, specs)
Implementation Clarity: 30 pts (FRs with codes, NFRs, architecture, phases)
Completeness: 20 pts (risks, dependencies, examples, doc quality)

PHASE 4: SAVE

Write to user-specified filename or generate based on feature name.

START WITH PHASE 0 CLARIFICATION QUESTIONS NOW.

Fluxly profile Author and license come from source; runtime, permissions, and network are Fluxly detections or estimates

Fluxly category: Documentation
Author-declared agents: No explicit declaration found; this is not inferred or tested compatibility
Static check: 88 / 100 · heuristic scan, not runtime safety proof
Author / version / license: @nyldn · no license declared
Fluxly token estimate: Lean
Fluxly setup estimate: Guided setup
External API key: No requirement detected
Detected OS requirements: Unspecified
Runtime requirements: Unspecified
Detected file/system behavior: Read-only
Write / modify
Shell exec
Detected network behavior: Local-only
Install commands: None (reference only)

Profile is derived at build time from SKILL.md and install vectors. Subject to drift from author intent.

Heads up: 未限定 allowed-tools，默认拥有全部工具权限。

Output preview skill-prd.preview

The current SKILL.md does not define a fixed output example.

SKILL.md SKILL.md

> **Host: Codex CLI** — This skill was designed for Claude Code and adapted for Codex.
> Cross-reference commands use installed skill names in Codex rather than `/octo:*` slash commands.
> Use the active Codex shell and subagent tools. Do not claim a provider, model, or host subagent is available until the current session exposes it.
> For host tool equivalents, see `skills/blocks/codex-host-adapter.md`.


# STOP - SKILL ALREADY LOADED

**DO NOT call Skill() again. DO NOT load any more skills. Execute directly.**


## PHASE 0: CLARIFICATION (MANDATORY)

Before writing ANY PRD content, you MUST ask the user these questions:

```
I need to understand your requirements before creating the PRD.

1. **Target Users**: Who will use this? (developers, end-users, admins, etc.)
2. **Core Problem**: What specific pain point does this solve? Any metrics?
3. **Success Criteria**: How will you measure if this succeeds?
4. **Constraints**: Any technical, budget, or timeline constraints?
5. **Existing Context**: Is this greenfield or integrating with existing systems?

Please answer these (even briefly) so I can create a more targeted PRD.
```

**WAIT for user response before proceeding to Phase 1.**

If user says "skip" or provides the feature description inline, extract what you can and note assumptions.


## PHASE 1: QUICK RESEARCH (Max 2 searches)

Only search if topic is unfamiliar. Limit to 2 web searches max:
- One for domain/market context
- One for technical patterns (if needed)

Do NOT over-research. 60 seconds max for this phase.


## PHASE 2: WRITE PRD

Structure:
1. **Executive Summary** - Vision + key value prop
2. **Problem Statement** - Quantified pain points by user segment
3. **Goals & Metrics** - SMART goals, P0/P1/P2 priority, success metrics table

…

Evidence boundary and execution chain

Author text anchors workflow facts; Fluxly only indexes current sections, terms, files, and commands.

sections -> PHASE 0: CLARIFICATION (MANDATORY) → PHASE 1: QUICK RESEARCH (Max 2 searches) → PHASE 2: WRITE PRD → PHASE 2.5: ADVERSARIAL PRD REVIEW (RECOMMENDED) → PHASE 3: SELF-SCORE → PHASE 4: SAVE
terms    -> Host: Codex CLI · DO NOT call Skill() again. DO NOT load any more skills. Execute directly. · Target Users · Core Problem · Success Criteria · Constraints · Existing Context · WAIT for user response before proceeding to Phase 1.
files/cmd -> /octo: · skills/blocks/codex-host-adapter.md · Adversarial review: applied · --fast · prd.md
body sha256 -> 590aacc52722