pulse-prd-review
Skill44 starsChanged 5 months agoApache-2.0View on GitHub
Use this skill when the user wants to review or validate an existing Pulse PRD against the execution framework, glossary, and folder conventions. Produces a structured findings report covering persona/layer correctness, metric reuse, hypothesis quality, dependency coverage, missing sections, front-matter compliance, and renderer-contract conformance. Triggers on phrases like "review this PRD", "validate this PRD", "audit PRD against framework", "does this PRD comply", "check PRD".
In this repository20 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- product/CLAUDE.mdCLAUDE.md
- pulse-web-otel/CLAUDE.mdCLAUDE.md
- add-ai-sub-agentSkill
- add-alert-metricSkill
- add-api-endpointSkill
- add-ui-componentSkill
- add-ui-screenSkill
- clickhouse-migrationSkill
- deploy-serviceSkill
- fatal-fiveSkill
- integrate-pulse-web-sdkSkill
- pr-reviewSkill
- pulse-prd-authorSkill
- web-otel-spec-audit-orchestratorSkill
- web-otel-spec-implementation-auditSkill
- web-sdk-e2e-matrixSkill
- web-sdk-instrumentSkill
- web-sdk-shipSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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 12
Read these files to load context.
- Distinctive: line 19
If the target PRD is unspecified, ask which file.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Pulse — PRD Reviewer
Audit an existing Pulse PRD against the framework and folder rules. Output is a structured review with explicit pass / soft / blocking findings — no prose drafting, no rewriting.
Before you start
Read these files to load context.
product/CLAUDE.md— folder rules, front-matter schema, renderer contract.product/frameworks/execution-framework.md— personas, layers, progress matrix, glossary.product/prds/_template.md— structural template to compare against.- The target PRD file the user named.
If the target PRD is unspecified, ask which file. Do not guess.
Review checklist
Walk every item. For each, mark ✅ Pass, ⚠️ Soft, or ❌ Blocking. Capture the specific line, section, or quoted phrase that triggered the finding.
Front-matter
- Front-matter block is present at the top.
title,status,layer,persona,last-edited,ownerare all present.layervalue is one of: detect / diagnose / quantify / resolve / predict / framework / meta.personavalue is one of: tech / product / ux / all (with body justification ifall).statusis one of: draft / in-review / approved / in-execution / live.last-editedisYYYY-MM-DD.trackeris present ifstatusisapproved,in-execution, orlive.
Anchoring
- Personas explicitly named in § Persona(s) Affected. If front-matter says
all, body has a reason. - Execution layer(s) explicitly named. Body matches front-matter.
- PRD locates the work in the progress matrix (mentions which cell it fills or upgrades).
Hypothesis
- Hypothesis section uses the standard shape: "We believe [behavior] for [persona] at [layer] will move [metric] because [reason]."
- The metric is from the glossary in
frameworks/execution-framework.md. A custom metric is a Soft finding unless flagged in Open Questions for glossary review — otherwise Blocking. - Behavior is observable, not philosophical.
Metrics & success
- Success Metrics section reuses glossary metrics where applicable.
- Each metric has Source, Baseline, Target columns filled (or marked as Open Question).
- No metric invented inline that should obviously map to a glossary entry.
Coverage
- Problem section has evidence (data point, customer quote, ticket reference, telemetry). "We think users are unhappy" is a Soft finding.
- Goals are concrete outcomes, not activities.
- Non-goals section exists and is non-empty.
- Dependencies cover all relevant areas: Backend / SDK / Pulse UI / Ingestion / AI Agent. If a clearly-touched area is missing, that is a Blocking finding for
in-reviewPRDs. - Out of Scope / Future Work present.
Renderer contract
- Status emojis (when used) are exactly ✅ ⚡ ❌. Substitutes (🟢 🟡 🔴, 🚧, ✓,
[x], "Done") are a Blocking finding. - Layer / Persona names use the exact capitalisation defined in the framework: Detect, Diagnose, Quantify, Resolve, Predict, Tech, Product, UX.
- Cross-doc links are relative (
../frameworks/...,../prds/...), not absolute paths or repo URLs.
Hygiene
- No duplicate definitions of personas, layers, or metrics — references point to the framework.
- No
TBD/???outside the Open Questions section. - No images, GIFs, or videos embedded — those break copy-paste.
Output format
Print the review using this exact structure (replace bracketed placeholders, omit empty sections):
# PRD Review — <PRD title>
**File:** `<path>`
**Reviewed against:** product/CLAUDE.md, product/frameworks/execution-framework.md
**Reviewed on:** <YYYY-MM-DD>
## Verdict
<APPROVED | APPROVED WITH NITS | NEEDS CHANGES | BLOCKED>
## Blocking findings (❌)
- <finding> — <line/section reference>
Recommendation: <action>
## Soft findings (⚠️)
- <finding> — <line/section reference>
Recommendation: <action>
## Passing checks (✅)
- Front-matter
- Anchoring
- Hypothesis
- Metrics
- Coverage
- Renderer contract
- Hygiene
## Suggested next step
<one sentence>Anti-patterns to refuse
- Re-drafting sections of the PRD inside the review. The review reports findings; it does not rewrite. If the user wants edits, they should pass the findings back through
/pulse-prd-authoror edit manually. - Soft-pedalling ❌ findings into ⚠️. The renderer contract and front-matter schema are non-negotiable.
- Approving a PRD with
status: approved/in-execution/livethat lacks atracker— that's a Blocking finding. - Auto-approving because "the doc looks fine." Walk the entire checklist before issuing a verdict.
When done
Tell the user:
- The verdict.
- The single highest-leverage fix to make next.
- Whether the PRD is ready to advance status (e.g.
draft → in-review).
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.

