Guide · Agent skill
How to use the Automation Audit Operations agent skill (automation-audit-ops)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The automation-audit-ops skill is a set of instructions that teaches an AI agent to audit the automations already in your workspace: local hooks, GitHub Actions, MCP server configs, connectors and wrapper scripts. It is for developers and teams who suspect they have duplicate or dead automation and want a clear list of what is actually configured, what is broken, and what overlaps. After following this guide you will know how to install the skill, how to phrase requests so the agent triggers it, and what kind of report to expect back.
Automation Audit Operationsaffaan-m/ECC · MIT Download the skill Quick start
- 1Download the ZIP. Download the automation-audit-ops package from this page. It contains the SKILL.md file.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use or into the project's .claude/skills/ folder for one repo.
- 3Or upload in the Claude apps. If you use the Claude apps, go to Settings → Capabilities → Skills and upload the package.
- 4Open the workspace. Start your agent session inside the repository you want audited so it can read hooks, workflows and MCP config.
- 5Ask a natural audit question. Ask which automations are live, broken or overlapping, in plain language.
- 6Review the evidence table. Check that each live-state claim cites a file path, log or config entry.
- 7Decide keep, merge, cut or fix next. Act on the recommendations, starting with the broken, high-signal items the report identifies.
What does the automation-audit-ops skill make the agent do?
Without the skill, a question like 'what automations do I have?' tends to get answered from memory or from whatever the agent last saw. With the skill installed, the agent treats that question as an audit task. It reads the actual setup first: repo hooks and hook scripts, GitHub Actions and scheduled workflows, MCP configs and enabled servers, connector-backed integrations, and wrapper scripts.
Each item is then sorted into one of five states: configured, authenticated, recently verified, stale or broken, or missing entirely. Important claims are backed by evidence such as file paths, workflow runs, hook logs, config entries or exact failure messages. The skill is explicit that a tool should not be called live just because a config or skill file mentions it.
- Reads the real setup rather than answering from memory
- Labels every surfaced automation with a live-state category
- Separates problem types: active breakage, auth stalls, stale state, overlap or redundancy, missing capability
- Ends each overlapping item with one call: keep, merge, cut or fix next
Install the skill from the ZIP download
Download the ZIP for this skill from this page. The package contains the SKILL.md file. You can install it in one of two ways depending on where you want it available.
For a personal install across your projects, unzip it into your user skills folder at ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder so it is available only there.
If you use the Claude apps rather than a local setup, you can upload the skill instead. Go to Settings, then Capabilities, then Skills, and upload the package there.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills, then upload
~/.claude/skills/Prepare the workspace you want audited
The skill reads the workspace it is pointed at, so the agent needs access to the repository, its hook scripts, its workflow files and its MCP configuration. Without that access the agent cannot inventory anything real.
The skill also refers to several sibling ECC skills: workspace-surface-audit, knowledge-ops, github-ops, ecc-tools-cost-audit, research-ops and verification-loop. These are not included in this file. The audit still works from the SKILL.md alone, but those referenced skills are useful when the work calls for them.
- Access to the repo, hook scripts, workflow files and MCP config
- Optional: access to CI logs or hook logs if you want live-state proof
- Optional: the sibling ECC skills named above, which are not part of this download
Trigger the skill with natural requests
You do not need a special command. Ask the question the way you would ask a colleague. The skill is meant to activate when you ask which automations exist, which are live, which are broken, or what overlaps.
The skill also applies when you are porting automations from another agent system and need to know what still has to be rebuilt in ECC.
- 'What automations do I have in this repo, and which of them are actually running?'
- 'Which of our GitHub Actions workflows are still scheduled and which look broken?'
- 'We have three MCP servers doing similar things. What overlaps and what should we keep?'
- 'I'm moving automations over from another agent setup. What is still missing in ECC?'
Read the audit output
The skill asks for a fixed output shape so that results are easy to compare. The report lists the current surface with its source and live state, then findings grouped by active breakage, overlap, stale status and missing capability. It finishes with a recommendation for each overlapping item and a next ECC move.
Where the state is unclear, the skill tells the agent to say so directly rather than present the audit as complete. Treat any entry without a proof path as unverified.
- CURRENT SURFACE: automation, source, live state, proof
- FINDINGS: active breakage, overlap, stale status, missing capability
- RECOMMENDATION: keep, merge, cut or fix next
- NEXT ECC MOVE: the skill, hook, workflow or app lane to strengthen
Safety notes: scripts and network
The skill itself does not run scripts and does not use the network. It starts read-only and makes no changes unless you explicitly ask for a fix. It also does not merge or delete overlapping automations until an evidence table exists.
Be aware that the audit may read hook scripts, workflow files and config that contain sensitive values. Review what the agent prints before sharing a report. If you want live proof, you must provide CI or hook logs yourself; the skill will not fabricate status from configuration alone.
- Runs scripts: no
- Uses network: no
- Read-only by default; changes only on explicit request
- Live status needs logs you supply; configuration alone is not proof
What people use it for
- Finding which scheduled GitHub Actions workflows still run and which have stopped working
- Consolidating several local hooks that repeat the same check into one canonical hook
- Deciding which of multiple MCP servers to keep when they expose overlapping tools
- Listing what still has to be rebuilt after porting automations from another agent system into ECC
Frequently asked questions
How do I install the automation-audit-ops skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
Is the automation-audit-ops skill free?
The skill is published under the MIT licence, as shown on this page. Check the repository's licence file for the full terms before redistributing it.
Does the skill require any other tools?
It needs the SKILL.md file and access to the workspace being audited. Sibling ECC skills such as workspace-surface-audit and github-ops are referenced but are not included in the download, and CI or hook logs are needed only if you want live-state proof.
Does the skill run scripts or use the network?
According to this page, it runs no scripts and uses no network. It works read-only by default and only makes changes when you ask for them.
Why does the agent say an automation is not live when the config shows it?
The skill is designed not to treat a configured or referenced tool as live. Configuration shows that something exists, while live status needs proof such as a workflow run or hook log.
How is this different from just asking the agent about my hooks?
Without the skill, the agent may answer from memory. With it, the agent reads the actual setup first, labels each item by live state and cites evidence before recommending keep, merge, cut or fix next.
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 your repo or team has accumulated overlapping hooks, workflows or MCP servers and you want one canonical setup backed by evidence. Pick something else if you need to build a new automation from scratch, since this skill deliberately avoids rewrites, or if you cannot give the agent access to logs or CI output and need proof of live status.
Download the skill