API Reference Audit Skill: What It Does and How to Install It

Audits a published API reference against its spec, endpoint by endpoint, then sets up the gates that stop docs drifting.

Installs: 9.7K; on skills.sh
Stars: 9; on GitHub
Files: 4; besides SKILL.md
Updated: Nov 2026; last repo commit

api-reference-quality makes Claude act as an API reference auditor. It builds an endpoint census from your spec, grades every operation against a rubric, scores the gaps by severity, and then helps you pick how to keep the reference accurate: a lint gate, a lint gate plus contract tests, or a full generated-docs migration.

It loads when you mention API reference docs, OpenAPI completeness, undocumented endpoints or error codes, or stale docs. It asks for your OpenAPI spec first; without one it starts from route definitions or gateway config. It covers endpoint-level reference only and bundles three reference files and no scripts.

Good for

API and docs engineers whose reference is out of date
Platform teams getting support tickets about undocumented parameters
Teams hand-writing reference pages who want a source-of-truth setup

Install api-reference-quality

Claude Code

Copy this prompt.

Install the api-reference-quality skill in Claude Code. Run this command:

/plugin marketplace add samber/developer-platform-skills
/plugin install developer-platform-skills@developer-platform-skills

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

The README installs every skill in the repo (61 skills/*/SKILL.md in the git tree), because skills here reference each other. api-reference-quality is one of them. Pick one route, not both.

From the project README. Source.

What's in the skill

When it loads

Audit a published API reference at the endpoint level against its spec surface - every operation, parameter, response code, and error documented, request/response examples present, per-language code snippets in parity, try-it affordances working - and install the source-of-truth gates that stop reference drift (spec-driven generation, OpenAPI lint, contract tests, snippet parity in CI). Use whenever the user mentions API reference docs, OpenAPI docs completeness, undocumented endpoints or error codes, stale or drifting docs, or docs CI gates - even if they never say "reference quality". Endpoint-level reference only. Do NOT use for portal IA and onboarding - use samber/developer-platform-skills@developer-portal-design instead.

Sections in SKILL.md

Clarifying questions
Reader types
Workflow
1. Census the surface
2. Audit every endpoint
3. Score and report
4. Choose the remediation tier
5. Run the staged rollout
6. Measure, gate, and re-audit
Failure modes
Invocation examples
References

Key facts

Author: Samuel Berthe
Repo: samber/developer-platform-skills, which holds 61 skills
License: MIT (LICENSE, README, plugin.json and SKILL.md all say MIT)
Version in SKILL.md: 1.0.2
Installs on skills.sh: 9,671 (Nov 2, 2026)
GitHub stars (repo): 9 (Nov 2, 2026)
Last commit (repo): Nov 1, 2026
What installs: 5 files including SKILL.md: 3 reference files and evals/evals.json; no scripts
Needs: The other skills in the repo; the README says to install all of them
Works in: Claude.ai, Claude Code, Codex, Cursor, Gemini CLI, and tools that read skills.sh installs

What it includes

Asks six questions, starting with whether you have a machine-readable spec.
Question 1 requests your machine-readable spec; no spec means the audit's first blocker finding.
Try: “Audit our API reference against this OpenAPI spec - I think half the error codes aren't documented.”
Clarifying questions
Builds an endpoint table and diffs spec against the published reference both ways.
Columns: operation, in spec, in published reference, in SDKs, discrepancy.
1. Census the surface
Grades each operation on parameters, responses, errors, examples and snippets.
Findings are graded blocker, major or minor using references/audit-rubric.md.
2. Audit every endpoint
Delivers coverage percentages and a per-endpoint findings table.
Also states whether hand-maintained docs mean every finding will recur.
3. Score and report
Ranks three tiers; the default is a lint gate plus contract tests.
Contract tests are dropped from the menu when there is no environment that can safely take real test requests.
4. Choose the remediation tier
Five ordered stages, from spec as source of truth to institutionalizing it.
Never recommends a later stage to a team that skipped an earlier one.
5. Run the staged rollout
Sets pass gates and a quarterly re-audit.
Census coverage of 100 percent is the skill's own definition of done.
6. Measure, gate and re-audit

From the official docs

Permissions

Medium risk

This skill writes audit advice and bundles no scripts, but it recommends contract tests that send requests to your API. Point any such tests at a sandbox, not production.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Contract tests send requests to your API
If you pick the contract-test tier, SKILL.md says it needs an environment where property-based tests can safely fire real requests. Point them at a sandbox, not production. github.com
It may save notes to developer-platform-context.md
SKILL.md says that when your agent keeps memory in a file, it should consider using this file name. Our advice: keep it out of commits if it holds company plans. github.com

Technical details

Scopes	Asks you clarifying questions and writes an audit report from your answers

Questions

How do I install api-reference-quality?

Use the install box at the top of this page. The README says to install every skill in samber/developer-platform-skills, not one, because the skills hand work to each other. Pick one route per agent.

Does api-reference-quality work in Claude, Codex and Cursor?

The README documents install routes for Claude.ai (plugin marketplace), Claude Code, Codex, Cursor, Gemini CLI and the skills.sh CLI. The install box lists each one.

What does it ask me before it designs anything?

SKILL.md lists six questions to ask first, and the answers change the later steps. Have your current setup and constraints ready.

Does it run scripts or need an API key?

It bundles no scripts. The folder holds SKILL.md, reference Markdown files and an evals file, and the skill works by asking you questions and writing advice.

Is api-reference-quality safe to install?

It bundles no scripts, so installing it adds instructions and reference files rather than programs. SKILL.md also suggests that, if your agent keeps notes in a file, it uses developer-platform-context.md. Our advice: keep that file out of public commits if it holds company plans, and check any figures the skill cites against their sources before quoting them to customers.

How is it different from developer-portal-design?

api-reference-quality audits endpoint-level reference against the spec. SKILL.md says not to use it for portal information architecture and onboarding, and points to developer-portal-design for that.

Does it call my API?

SKILL.md tells the agent to take your spec as the census source and, where feasible, to look for undocumented surface in gateway logs, SDK internals and support tickets. Contract tests, if you choose that tier, are the step that sends requests; SKILL.md says they need an environment where that is safe.

Information

updated Nov 2, 2026

Repository: samber/developer-platform-skills
Folder: skills/api-reference-quality

samber/developer-platform-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

samber/developer-platform-skills README · github.com
api-reference-quality SKILL.md · github.com