← Back to Skills Marketplace
allens0104

Adapter Audit

by AllenS0104 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
81
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install adapter-audit
Description
Use this skill to audit CLI adapter projects (like opencli) for missing output fields, then batch-generate fixes and submit PRs. Turns AI agents into adapter...
README (SKILL.md)

Adapter audit skill

Use this skill when you want to systematically audit and fix CLI adapter projects — for example, scanning all search adapters in opencli for missing url fields and batch-generating patches.

When to use

  • auditing adapter output schemas for missing fields (url, timestamp, thumbnail, etc.)
  • generating batch fixes across many adapters at once
  • preparing multi-file PRs to upstream projects
  • maintaining output consistency across a large adapter registry

Core workflow

1. Scan

Scan all adapters in the target project and classify each by:

  • adapter type (YAML declarative vs TS/JS coded)
  • which output fields are present
  • which standard fields are missing

Standard fields to check:

  • url — direct link to the content item
  • title — content title
  • author — creator/author name
  • timestamp / date — publish or capture time

2. Classify fix type

For each missing field, determine the fix strategy:

Situation Fix strategy
Field is computed internally but not in columns Add to columns list
Field is computed but stripped by a map step Pass through the map step + add to columns
Field is not computed but can be constructed from existing data Add construction logic + add to columns
Field requires external data not available in the API response Skip or mark as "needs upstream API change"

3. Fix

Apply fixes using the minimum change principle:

  • For YAML adapters: add field to return object + map step + columns array
  • For TS adapters: add field to return object + columns array
  • Never change existing fields or behavior
  • Only add new fields

4. Verify

After fixing, run the project's existing tests:

npm run build    # ensure TS compiles
npm test         # ensure nothing breaks

5. Submit

Create a single well-documented PR with:

  • clear title describing the scope
  • table showing before/after coverage
  • per-adapter fix type classification
  • risk assessment (should always be "additive only")

Proven example

This skill was used to audit opencli's 33 search adapters:

  • Before: 22/33 (67%) had url in output
  • After: 32/33 (97%) had url in output
  • Fix types used: columns-only (3), map-passthrough (2), construct-from-data (4)
  • Files changed: 9
  • Lines changed: +17 / -10
  • PR: merged within hours, all CI green

Output contract

The audit output should include:

  1. total adapters scanned
  2. per-adapter field coverage table
  3. fix strategy for each missing field
  4. list of files changed
  5. before/after coverage metrics
  6. risk classification

Key rules

  • Never modify existing output fields — only add missing ones
  • Prefer constructing URLs from existing API data over adding new API calls
  • Skip fields that genuinely don't apply (e.g., url for dictionary word lookup)
  • Always verify the constructed URL format is correct for the platform
  • Group all fixes into a single PR for easier review

Quick invocation template

Use /adapter-audit to scan all search adapters in this CLI project for missing url fields, fix them, and prepare a PR.
请用 /adapter-audit 扫描这个 CLI 项目里所有 search adapter 的缺失字段,批量修复并准备 PR。
Usage Guidance
This skill is instruction-only and seems designed to edit your repository and create PRs, but the package metadata omits practical requirements. Before installing or running it: 1) Treat it as a tool that will modify your workspace — run it on a fork or disposable branch. 2) Expect it to need git, node/npm (to run npm run build/test), and a GitHub auth method (GITHUB_TOKEN or SSH key) to push and open PRs; ask the publisher which credentials are required and how they are used. 3) Provide least-privilege credentials (e.g., a token scoped only to repo creation/PR for a specific repo or use a fork workflow). 4) Confirm whether the agent will push to upstream or to your fork and whether it will force-push. 5) If you cannot safely provide push credentials, consider running the audit in 'dry-run' mode: have the agent generate patch files or a local branch only, then review and push manually. 6) Ask the publisher to update the skill metadata to declare required binaries and environment variables and to document authentication and PR target behavior. If you cannot get these clarifications, treat the skill as potentially risky and prefer manual or sandboxed execution.
Capability Assessment
Purpose & Capability
The SKILL.md describes scanning, modifying adapters, running build/tests, and submitting a PR — all coherent with an 'adapter-audit' purpose. However the registry metadata claims no required binaries or env vars, which is inconsistent: practical execution requires tools like git, node/npm (for npm run build/test), and a mechanism/credential to push branches and open PRs (e.g., GITHUB_TOKEN).
Instruction Scope
Instructions stay on-topic (scan repository files, add fields, run build/test, create a PR). They do not instruct reading unrelated system files or exfiltrating data. Missing are explicit, deterministic steps for authentication and for whether changes should be pushed to a fork or upstream — giving the agent broad discretion about push/PR targets.
Install Mechanism
This is an instruction-only skill with no install spec or code files; that reduces installer risk because nothing is written by the skill package itself.
Credentials
No credentials or environment variables are declared even though the skill's primary function (submitting PRs) typically requires repository write access or a GitHub token. The absence of declared required auth is a mismatch and could result in the agent attempting to use ambient credentials or prompting the user for secrets at runtime.
Persistence & Privilege
The skill is not always-enabled. It will modify local repository files and run build/test commands in the user's workspace — normal for this purpose, but potentially impactful. Because the agent can act autonomously (platform default), if the agent has push privileges the skill could create branches/PRs; the skill itself does not request persistent privileges or alter other skills.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install adapter-audit
  3. After installation, invoke the skill by name or use /adapter-audit
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of adapter-audit skill. Key features: - Audits CLI adapter projects for missing standard output fields (url, title, author, timestamp). - Scans and classifies adapters, suggests and applies minimal batch fixes. - Prepares multi-file PRs with detailed before/after coverage and risk assessment. - Ensures only additive changes; never modifies existing output behavior. - Includes step-by-step workflow and usage instructions.
Metadata
Slug adapter-audit
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Adapter Audit?

Use this skill to audit CLI adapter projects (like opencli) for missing output fields, then batch-generate fixes and submit PRs. Turns AI agents into adapter... It is an AI Agent Skill for Claude Code / OpenClaw, with 81 downloads so far.

How do I install Adapter Audit?

Run "/install adapter-audit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Adapter Audit free?

Yes, Adapter Audit is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Adapter Audit support?

Adapter Audit is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Adapter Audit?

It is built and maintained by AllenS0104 (@allens0104); the current version is v1.0.0.

💬 Comments