agentleFS
Sign inStart free
kv0906kv0906/pm-kit/.claude/skills/doc/SKILL.md

doc

Skill138 starsChanged 7 months agoMITView on GitHub

Draft or edit a document — PRD, spec, design doc, technical brief. Use for "/doc project title" to create a new doc or edit existing.

We split this skill into the rules it gives an agent

13 rules. 0 are common, 8 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 26 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create or edit documents (specs, PRDs, design docs, etc.).

    line 10
  2. Very specific:

    Search docs/{project}/*.md for similar titles (or docs/general/*.md if no project)

    line 34
  3. Distinctive:

    If found: Offer to edit or create new

    line 35
  4. Distinctive:

    Read the user's profile and relevant project index.

    line 38
  5. Distinctive:

    Search related docs, customer feedback, meetings, decisions, and blockers

    line 39
  6. Very specific:

    Use /discover when the problem needs investigation and /prioritize when competing requests need a scope decision

    line 42
  7. Very specific:

    Follow _core/judgment.md to select outcome measures.

    line 48
  8. Distinctive:

    Do not invent metric values, customer quotes, priorities, owners, or technical decisions

    line 51
  9. Distinctive:

    Keep technical approach optional when unknown.

    line 54
  10. Very specific:

    If no project specified: write to docs/general/{slug}.md

    line 57
  11. Very specific:

    If project specified: write to docs/{project}/{slug}.md

    line 58
  12. Distinctive:

    When editing, preserve existing content and structure.

    line 63
  13. Distinctive:

    Add to project index if exists

    line 67

“Common” is measured: kept by at least 6 in 10 of the 26 edited versions of this file we could read. “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

/doc — Document Drafting

Create or edit documents (specs, PRDs, design docs, etc.).

Context

Today's date: !date +%Y-%m-%d Existing docs: !ls docs/*/*.md 2>/dev/null | tail -10

Reference template: @_templates/doc.md Config: @_core/config.yaml Processing logic: @_core/PROCESSING.md Product judgment: @_core/judgment.md Knowledge and evidence: @_core/knowledge.md

Input

User input: $ARGUMENTS

Processing Steps

  1. Parse Input
    • Extract project
    • Extract title/feature name
  2. Check Existing
    • Search docs/{project}/*.md for similar titles (or docs/general/*.md if no project)
    • If found: Offer to edit or create new
  3. Gather Context
    • Read the user's profile and relevant project index; follow linked product context and approved judgment/writing examples
    • Search related docs, customer feedback, meetings, decisions, and blockers
    • Identify the affected user, their task, observed problem, and desired business outcome before proposing requirements
    • Separate sources, interpretations, and assumptions; retain conflicting evidence and mark missing evidence as unknown
    • Use /discover when the problem needs investigation and /prioritize when competing requests need a scope decision
  4. Interactive Drafting
    • Accept existing content; ask only for gaps that materially affect the problem, scope, or validation
    • Draft problem statement, evidence, assumptions, goals/non-goals, and requirements at the depth the task needs
    • For goals, state the intended outcome; include metric definition, baseline, target, period, and source when known
    • Follow _core/judgment.md to select outcome measures; reuse existing goals rather than forcing a North Star, OKR, or metric tree
    • For consequential releases or experiments, include relevant guardrails, instrumentation gaps, and proposed scale/iterate/stop criteria
    • Propose the smallest validation step and review point; distinguish proposed targets from approved commitments
    • Do not invent metric values, customer quotes, priorities, owners, or technical decisions
    • For an approved feature direction, break work into small user stories with actor, need, benefit, and observable acceptance criteria
    • Cover relevant failure and recovery cases; flag dependencies. If direction is unapproved, label any stories as proposed
    • Keep technical approach optional when unknown; match the requested audience, language, and approved voice examples
  5. Create Doc Note
    • If no project specified: write to docs/general/{slug}.md
    • If project specified: write to docs/{project}/{slug}.md
    • Apply template with gathered data
    • For saved explanations, research briefs, comparisons, or journey maps, adapt template sections to the purpose; do not add PRD requirements to a thinking note
    • Status: draft
    • Mark missing sections as unknown or not applicable; never leave unresolved template variables
    • When editing, preserve existing content and structure; add evidence and validation only where they are relevant
    • Add ## Links section
  6. Update References
    • Add to project index if exists
  7. Append to Vault Log
    • Append entry to 01-index/_vault-log.md (see .claude/rules/vault-log.md)
    • Action: doc
    • Details: doc title and status

Output

Created: docs/{project}/{slug}.md
Status: draft
Sections: {filled}/{total}

Highlight important evidence gaps and the next validation step. Creating a local draft does not authorize publishing it or creating external tickets.

Export (Optional)

Supports --docx, --pdf flags. See .claude/rules/export-formats.md for layout specs and workflow. Complete normal processing first, then generate the formatted file.

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.

Check your setup, free