agentleFS
richyuhrichyuh/claude-skills/customer-facing-spec/SKILL.md

customer-facing-spec

Skill0 starsChanged 2 months agoMITView on GitHub

Write a customer- and partner-facing technical/product specification that describes product capabilities, user-facing behavior, integration requirements, limits, and compliance posture — while strictly excluding proprietary internals (architecture, infrastructure, vendors, algorithms, roadmap, pricing internals). Use when asked to draft, write, or review an external product spec, partner spec, customer-facing technical specification, or a public product/feature description for customers or partners. Includes a mandatory pre-publish leak-check pass.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Strict allow-list redaction: Only information that is explicitly customer/partner-relevant is included.

    line 13
  2. Distinctive:

    Ask 3–6 focused questions if scope is unclear.

    line 20
  3. Very specific:

    Use reference/spec-template.md as the structure.

    line 23
  4. Very specific:

    Before presenting the finished spec, perform an explicit second pass dedicated solely to leakage.

    line 29
  5. Distinctive:

    Never present a spec as final until the Leak-Check Report has been shown and the ⚠️/❓ items resolved with the user.

    line 34
  6. Distinctive:

    When unsure whether something is external-safe, treat it as proprietary and ask.

    line 43
  7. Very specific:

    Never substitute a real internal vendor/tool/system name with a plausible-but-invented one.

    line 44
  8. Distinctive:

    Do not infer or speculate about roadmap, pricing internals, or unreleased capabilities.

    line 45

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Customer-Facing Product Specification

Produce an external narrative product specification for customers and partners. The reader is outside the company (a customer evaluating the product, or a partner integrating with it). They need to understand what the product does, how it behaves, and how to work with it — never how it is built internally.

Two non-negotiable principles govern everything below:

  1. Describe behavior, not implementation. State observable product behavior, user-facing contracts, and partner requirements. Never state internal architecture, infrastructure, source systems, vendors, or algorithms.
  2. Strict allow-list redaction. Only information that is explicitly customer/partner-relevant is included. Anything that even resembles a proprietary internal detail is excluded by default and surfaced to the user as a question — not silently published. See reference/redaction-rules.md.

Workflow

1. Gather inputs (interview, don't assume)

Ask the user for: the product/feature name, the target audience (which customers/partners), the spec's purpose (evaluation, integration, contractual reference), and any source material (internal docs, PRDs, code). If source material contains internal detail, that's expected — your job is to translate the external-relevant parts and drop the rest.

Ask 3–6 focused questions if scope is unclear. Do not pad with assumptions.

2. Draft against the template

Use reference/spec-template.md as the structure. Fill only sections that apply; omit sections that don't rather than inventing content. Write in clear, plain prose aimed at a technical-but-external reader. Prefer concrete, verifiable statements over marketing language.

3. Apply the redaction rules continuously

While drafting, run every candidate statement through the allow-list test in reference/redaction-rules.md. When you encounter information that is useful but proprietary, do not include a sanitized guess — instead leave a [REDACTED — confirm external-safe phrasing with user] marker and list it for review in step 4.

4. Mandatory pre-publish leak-check (do not skip)

Before presenting the finished spec, perform an explicit second pass dedicated solely to leakage. Read reference/redaction-rules.md and check the draft line by line against the forbidden-category checklist there. Produce a short Leak-Check Report for the user containing:

  • ✅ Categories confirmed clean.
  • ⚠️ Every borderline statement, quoted verbatim, with the category it risks and a proposed external-safe rewrite.
  • ❓ Every [REDACTED …] marker that needs a user decision.

Never present a spec as final until the Leak-Check Report has been shown and the ⚠️/❓ items resolved with the user.

5. Editorial pass (after leak-check only)

Once the Leak-Check Report's ⚠️/❓ items are resolved, apply the writing-clearly-and-concisely skill to tighten grammar, structure, and clarity. This pass runs after redaction, never before. It must not rewrite, paraphrase, or "improve" any [REDACTED …] marker or reintroduce excluded detail — if a sentence can't be improved without touching redacted content, leave it as-is and note it.

6. Deliver

Output the spec as Markdown. If the user wants it published, suggest the official notion, atlassian (Confluence), or mintlify integrations — do not push it anywhere without explicit confirmation.

Hard rules

  • When unsure whether something is external-safe, treat it as proprietary and ask. Default to exclusion.
  • Never substitute a real internal vendor/tool/system name with a plausible-but-invented one; redact and flag instead.
  • Do not infer or speculate about roadmap, pricing internals, or unreleased capabilities.

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.