agentleFS
dream-horizon-orgdream-horizon-org/pulse/.agents/skills/pulse-prd-author/SKILL.md

pulse-prd-author

Skill44 starsChanged 5 months agoApache-2.0View on GitHub

Use this skill when the user wants to draft a new Pulse PRD from a feature brief, idea, or scoping conversation. Walks the user through locating the work in the execution-framework progress matrix, anchoring persona and execution layer, building a hypothesis using glossary metrics, and producing a draft PRD saved to product/prds/. Triggers on phrases like "write a PRD", "draft a PRD for X", "scope feature Y for Pulse", "create PRD for", "let's PRD this".

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read these files to load context.

    line 12
  2. Distinctive:

    If any of these files are missing, stop and tell the user before continuing.

    line 18
  3. Distinctive:

    Ask: *what cell of the progress matrix does this work fill?*

    line 26
  4. Very specific:

    Read frameworks/execution-framework.md § Progress Matrix.

    line 28
  5. Very specific:

    Add front-matter at the top per product/CLAUDE.md schema:

    line 72

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Pulse — PRD Author

Guided authoring of a new Pulse PRD. This is a brainstorm with checkpoints, not autonomous drafting. Stop and confirm with the user at every step.

Before you start

Read these files to load context. Do not skip — the rest of the skill assumes the rules they encode.

  1. product/CLAUDE.md — folder rules, front-matter schema, renderer contract.
  2. product/frameworks/execution-framework.md — personas, layers, progress matrix, glossary.
  3. product/prds/_template.md — the structural template the output must follow.

If any of these files are missing, stop and tell the user before continuing.

Workflow — five checkpoints

Walk these in order. Do not skip ahead. Do not write the PRD file until checkpoint 5.

Step 1 — Locate

Ask: what cell of the progress matrix does this work fill?

  • Read frameworks/execution-framework.md § Progress Matrix.
  • Identify which (Stage, Persona) cells the work touches.
  • Cross-check whether existing tools listed in those cells already cover the proposed work. If they do, surface this — the work might be an extension of an existing tool, not a new PRD.
  • Confirm with the user before proceeding.

Step 2 — Anchor

Lock down two things explicitly:

  • Persona(s) affected. Tech / Product / UX. Reject "all" unless the user provides a specific reason. Default expectation: one or two personas with a primary.
  • Execution layer(s). Detect / Diagnose / Quantify / Resolve / Predict. If the work spans layers, identify the dependency direction (e.g. "Diagnose tool needs a new Detect signal first").

Step 3 — Hypothesise

Build the hypothesis using this exact shape:

We believe [behavior] for [persona] at [layer] will move [metric] because [reason].

Constraints:

  • The metric must come from the glossary in frameworks/execution-framework.md § Metrics & Glossary. If the user proposes a new metric, push back: ask which existing metric is closest and why it doesn't fit. Only proceed with a new metric if the user has a clear answer — and flag it in Open Questions as requiring a glossary update.
  • The behavior must be observable, not philosophical. "Reduce diagnose effort" is bad. "Drop MTTDx from 45 minutes to under 10 minutes for Product persona on funnel anomalies" is good.

Step 4 — Scope

Pull from the user, in this order:

  • Problem — with evidence (data point, customer quote, support ticket, usage telemetry). No hand-waving.
  • Goals — concrete outcomes, not activities.
  • Non-goals — what we explicitly will not do, and why.
  • Solution overview — one paragraph, no implementation detail.
  • Dependencies — by area: Backend / SDK / Ingestion / Pulse UI / AI Agent. Cross-check the codebase by reading the repo's root CLAUDE.md for layout, and surface anything the user missed.
  • Out of scope / future work — what's deferred and the rough trigger to pick it up later.
  • Open questions — the parts the user can't answer yet.

Push back if any of these are vague. "TBD" lands in Open Questions, not in Goals.

Step 5 — Draft

Now produce the file:

  1. Slugify the feature name into kebab-case: <feature-slug>.
  2. Read product/prds/_template.md.
  3. Fill the template using the information gathered. Preserve the template's section order and headings.
  4. Add front-matter at the top per product/CLAUDE.md schema:
    • title — explicit.
    • status: draft
    • layer: and persona: from Step 2.
    • last-edited: today's date.
    • owner: — ask the user if not already known.
    • tracker: if available; otherwise leave blank and add to Open Questions.
    • hero: flat
  5. Save to product/prds/<feature-slug>.md.
  6. Show the user the saved path. List which sections still have TBD or open questions, so they know what to follow up on.

Anti-patterns to refuse

  • Writing the full PRD before reaching Step 5. Steps 1–4 are intake; only Step 5 produces the file.
  • Inventing a new metric. Always reuse the glossary unless the user explicitly justifies adding one.
  • Skipping the matrix locate step. "Where in the framework does this live?" is non-negotiable.
  • Auto-completing fields the user hasn't given evidence for. Better to leave a TBD and list it in Open Questions.
  • Substituting [x] / 🟢 / "Done" for the status emojis ✅ ⚡ ❌ defined in product/CLAUDE.md.

When done

Tell the user:

  • The path the PRD was saved to.
  • The open questions that need follow-up.
  • That /pulse-prd-review is the next step once Open Questions are resolved.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.