prd-draft-assistant
Skill19 starsChanged 50 days agoMITView on GitHub
Draft one section of a product requirements document (PRD) from notes, a ticket, or a confirmed brief, and write it to a file only with explicit approval. Use when requirements are already reasonably clear and need to be written up; do not use on a vague or conflicting request — use the Requirements Gap Investigator first.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- content-brief-gap-investigatorSkill
- content-draft-assistantSkill
- content-publish-readiness-checklistSkill
- feature-launch-readiness-checklistSkill
- requirements-gap-investigatorSkill
- post-performance-investigatorSkill
- social-caption-drafterSkill
- social-publish-readiness-checklistSkill
- changelog-entry-drafterSkill
- pre-merge-readiness-checklistSkill
- root-cause-investigatorSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 11 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
If the source material is vague or contains conflicting signals, say so and recommend the Requirements Gap Investigator instead of drafting …
- Distinctive: line 15
Never treat access to write a file as permission to write it.
- Distinctive: line 23
If the source material leaves a needed section un-fillable (e.g.
- Distinctive: line 27
Use standard PRD sections as needed — see references/prd-section-format.md — such as Problem Statement, Goals, Non-Goals, User Stories, …
- Distinctive: line 29
Keep language concrete and checkable — prefer "reduces average export time from 40s to under 5s" over "makes exports faster," when the …
- Distinctive: line 37
Default: one-time approval.
- Distinctive: line 39
Never infer standing approval from a single approval of one section.
- Distinctive: line 43
Write the approved section(s) to the named file, in the right place if the file already has other sections.
- Distinctive: line 47
Stop and ask rather than guessing when:
- Distinctive: line 56
Before writing, confirm that:
- Distinctive: line 58
every drafted statement traces back to the supplied source material.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
PRD Draft Assistant
Draft one PRD (or a single section of one) from material the user supplies, show it before writing, and write only with the user's approval.
Operating boundary
- Touch only the single PRD file named or clearly implied by the user (e.g.
PRD.md, a named doc, or a specific section within one). Do not edit unrelated files. - Base every drafted statement on the supplied notes, ticket, or brief. Do not invent requirements, metrics, or scope not present in the source material.
- If the source material is vague or contains conflicting signals, say so and recommend the Requirements Gap Investigator instead of drafting around the ambiguity.
- Never treat access to write a file as permission to write it. Always get approval first (see below).
Workflow
1. Gather the source material
Identify what you're drafting from: a ticket, a set of notes, a confirmed brief, or a prior gap-investigation report. State which one you used.
If the source material leaves a needed section un-fillable (e.g. no success metric was ever stated), say so explicitly in the draft rather than inventing a plausible-sounding one.
2. Draft the section(s)
Use standard PRD sections as needed — see references/prd-section-format.md — such as Problem Statement, Goals, Non-Goals, User Stories, Success Metrics, and Open Questions. Only draft sections the source material actually supports; leave others explicitly marked as needing input rather than filling them with plausible-sounding filler.
Keep language concrete and checkable — prefer "reduces average export time from 40s to under 5s" over "makes exports faster," when the source material supports a specific number; otherwise state it as an open question rather than inventing a number.
3. Show the draft and get approval
Show the complete drafted section(s) to the user before writing anything. Do not write first and confirm after.
Approval model:
- Default: one-time approval. Ask before this specific write; a "yes" or "go ahead" authorises this write only.
- Session-scoped approval: only if the user's own request explicitly grants standing permission for the session (for example, "draft each section as we go and save them without asking each time"). When this applies, say so out loud before proceeding, and continue asking again in any later session.
- Never infer standing approval from a single approval of one section.
4. Write and confirm
Write the approved section(s) to the named file, in the right place if the file already has other sections. Confirm what was written and where.
Stopping conditions
Stop and ask rather than guessing when:
- the source material is too vague or conflicting to draft from confidently — recommend the Requirements Gap Investigator instead;
- a section would require inventing a metric, scope boundary, or user story not present in the source; state it as an open question instead;
- the target file or section is ambiguous; or
- the request also asks for a product decision (e.g. prioritisation, go/no-go) — that is out of scope for this skill.
Quality check
Before writing, confirm that:
- every drafted statement traces back to the supplied source material;
- sections that can't be responsibly filled are marked as open questions, not invented;
- the draft was shown to the user before any write occurred; and
- the approval obtained matches what is being written (one-time vs. session-scoped, and for this write specifically).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

