| name | doc-prd |
| description | Synthesise the current conversation into a PRD and publish it to the issue tracker without re-interviewing — knowledge is assumed in context from a prior brainstorm or doc-align. Forbids file paths and code in the body (except prototype snippets that encode a decision). Skips to ready-for- agent.
|
PRD Synthesis Protocol
Overview
doc-prd exists to bottle a conversation. The user has already
brainstormed (via superpowers:brainstorming) or grilled (via
doc-align) or both, and the decisions live in the current context
window. This skill packages those decisions as a PRD, publishes it,
and labels it ready-for-agent so an AFK agent can pick it up.
It does not re-interview. If the knowledge is not already in
context, refuse and hand back to doc-align or
superpowers:brainstorming.
When to Engage
Engage when:
- A brainstorm or grilling session just concluded and decisions are
fresh in context
- The user says "PRD this", "turn this into a ticket", "write the
spec from what we just discussed"
Do not engage when:
- The conversation is exploratory and decisions are not pinned down —
grill first
- The user wants a synchronous ADR (use
/octopus:doc-adr)
- The output is a design doc with rationale (use
/octopus:doc-design
or /octopus:doc-spec)
Protocol
Step 1 — Explore the repo + glossary
Read CONTEXT.md and the relevant docs/adr/*. The PRD must speak
the project's vocabulary.
Step 2 — Draft the module sketch
Before writing the PRD body, list modules to be built or modified and
validate with the user. Actively look for deep modules — if the
sketch reveals three new shallow files, push back. Ask which modules
deserve test coverage.
This is the last interactive step. Everything after is synthesis.
Step 3 — Write the PRD body
Sections, in order:
- Problem — one paragraph, in domain language
- Solution — one paragraph, no implementation detail
- User Stories — exhaustive, not representative. Every actor /
path the conversation surfaced gets a story
- Implementation Decisions — the choices already pinned. Cite
ADRs where they apply
- Testing Decisions — which modules get tests, at what level,
and why
- Out of Scope — what was explicitly excluded
- Further Notes — open questions the agent should ask before
starting
Step 4 — Publish and label
Publish to the project's tracker (Notion / GitHub Issues / etc — the
publication layer is shared with doc-rfc). Apply ready-for-agent
directly — no re-triage round.
Forbidden Content
PRDs created by this skill must not contain:
- File paths — they rot when the codebase moves
- Code snippets — same reason, with one exception
The exception: a snippet from a prototype run that encodes a
decision more precisely than prose can — state machines, reducer
shapes, schemas. If a prototype produced a 12-line state machine that
captures the design exactly, include it. Cite the prototype.
Anti-Patterns
- Re-interviewing the user — if knowledge is missing, refuse
- File paths in the PRD body
- Code snippets that are not prototype-derived decisions
- Representative user stories ("e.g. a teacher does X") instead of
exhaustive ones
- Stopping at
needs-triage — the PRD is born ready-for-agent
- "TBD" markers in Implementation Decisions — if it is TBD, grill
more before publishing
Integration with Other Skills
doc-align — runs before doc-prd when decisions are not
yet pinned
superpowers:brainstorming — runs before doc-align when the
idea is still being shaped
prototype — produces the snippets allowed under the
exception in "Forbidden Content"
doc-rfc / doc-spec — sibling artifacts. PRD targets the
tracker for AFK agents; RFC/Spec target docs/ for design review
triage-issues — does not re-triage the published PRD