Guide · Agent skill
How to use the Maintain Verification Skill agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The maintain-verification-skill is an agent skill for keeping a project's verification skill and its product and agent access maps accurate against the current source code and real evidence. It is written for the Paperclip codebase and is aimed at maintainers and developers who want an honest audit of what is covered, what has changed, and what gaps remain. After this guide you will know what the skill makes an agent do, how to install it, how to trigger it, and what it needs to run safely.
Maintain Verification Skillpaperclipai/paperclip · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it into ~/.claude/skills/ or the project's .claude/skills/ folder.
- 2Confirm prerequisites. Make sure Node.js and Git are available, and that a Paperclip instance can run if you want the live pass.
- 3Ask for an audit. Invoke /maintain-verification-skill or describe a targeted audit, naming the areas you want covered.
- 4Run the structure check. The agent runs node scripts/check-maps.mjs to check map structure and indexes.
- 5Review source and live coverage. Read the agent's report, which lists source scope, live scope and remaining gaps separately.
- 6Review proposed corrections. Re-read any changed skill, map or check files before accepting them.
What is an agent skill?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent can load when a request matches the skill's description. This skill's SKILL.md describes a maintenance playbook: the steps an agent follows when auditing a verification skill.
What does the skill make the agent do?
When asked, the agent audits an existing project-local verification skill and its feature and agent-access maps against the current source code. It fixes outdated references or recipes only when it can prove they are wrong, then reports what is covered, what changed, and what gaps remain.
The playbook tells the agent to read source routes and permissions, run the repo's map structure check, and drive the running app through the existing verification workflow. It also keeps product code and permission rules off-limits for edits during a maintenance run.
- Locates the target verification skill, usually under .agents/skills/verify-*/
- Checks both map indexes for orphan or missing entries and broken references
- Reads source in read-only review passes, separating source review from executed proof
- Reports an outcome of clean, changed, or blocked, with source scope, live scope and remaining gaps listed separately
Install the maintain-verification-skill
Download the ZIP from this page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Project: unzip into the project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload
How do you trigger the skill?
You can invoke it directly with /maintain-verification-skill, or describe what you want in plain language. The skill's description covers auditing the verify skill and reconciling feature and agent access coverage, so requests that match those tasks should load it.
- /maintain-verification-skill
- Audit the verify skill for this project against the current source.
- Reconcile feature and agent access coverage for the billing area.
- Check whether every business operation in the feature map has an API and a permitted agent path.
What does the skill need before it runs?
The skill assumes a project-local verification skill already exists. If none exists, the skill points you to /create-verification-skill instead. The source lists these requirements for the full workflow.
- Node.js, to run scripts/check-maps.mjs
- Git, to inspect diffs and history
- A running Paperclip instance and its existing API, runner and MCP fixtures for the live pass
- Agent credentials for permitted agent identities during access checks
node scripts/check-maps.mjsSafety notes: scripts and network
This skill runs scripts and uses the network. The map structure check is a script run from the repo, and the live pass drives a running app and uses agent credentials.
The check script verifies structure only. It does not confirm API authorization, real tool bindings or live behavior. A permission denial is not permission to widen access, and the skill does not change product APIs or permission rules during a maintenance run. Use disposable fixtures for mutation and denial tests.
- Structure check does not prove authorization or live behavior
- Live tests should use disposable fixtures
- Access-policy changes are recorded as gaps for separate implementation
What people use it for
- Keeping a verify skill and its feature maps current after a batch of code changes
- Checking that each business operation has an API and a permitted agent path, or an explicit gap
- Producing a coverage report that separates source review from live proof
- Reconciling agent access companions with the feature recipes they support
Frequently asked questions
How do I install the maintain-verification-skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Is maintain-verification-skill free to use?
The source identifies its licence as MIT. Check the NOTICE.md file and the repository for the full licence text and attribution.
What do I need to run maintain-verification-skill?
You need Node.js to run scripts/check-maps.mjs and Git to inspect diffs and history. The live pass also needs a running Paperclip instance, its existing API, runner and MCP fixtures, and agent credentials.
Does this skill work outside the Paperclip codebase?
Its conventions are defined for Paperclip. For other projects, the skill maintains their existing verification skill and map locations and does not relocate them.
What if I don't have a verification skill yet?
The skill assumes one already exists. If none is found, it points you to /create-verification-skill, which is a separate skill for creating one.
Will the skill change my product APIs or permissions?
No. The skill forbids editing product APIs, permission enforcement, tool bindings or UI behavior during a maintenance run. It records gaps for separate implementation instead.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you maintain a Paperclip verification skill and its product and access maps and want an honest audit after code changes. Pick something else if you have no existing verification skill, need the skill to change product APIs or permissions, or work on a project with different map locations.
Download the skill