Works with the AI Agents you already use
Repo Vetter: Audit Any GitHub Repo Before You Run It
Audit untrusted repositories in a sandbox to provide a two-axis safety verdict before you clone or run code.
$12
Repo Vetter: Audit Any GitHub Repo Before You Run It
Example session with this skill installed
Vet this repo: https://github.com/example-user/quick-scrape. I found it on a Discord server and want to run it to extract some web data.
- Read your context and instructions
- Compiled the repo vetter
Verdict
To read (clone it to your machine and look at it): DO NOT CLONE
To run (via pip install quick-scrape): DO NOT RUN
Repo: example-user/quick-scrape — claims to: a lightweight CLI for asynchronous web scraping.
Trust signals: account created 14 days ago, 12 stars, single contributor, not a fork. PyPI quick-scrape published by the same account 9 days ago.
Findings:
setup.py:12→ custom install command downloads a script from a raw gist and executes it duringpip install. Code runs on your machine before you ever call the tool.requirements.txt→ depends onrequets(one letter offrequests), a known typosquat pattern; the PyPI package of that name is 6 days old with no linked repo.src/core.py:84→ hardcoded Discord webhook; posts hostname, username and the scrape target on every run. The README says nothing about telemetry, and a scraper has no reason to phone home.
Not checked: the gist's current contents (it can change at any time), therequetspackage's own code, and any dynamic run.
If you run it anyway: don't. Three independent deception signals — install-time remote execution, a typosquatted dependency, undisclosed exfiltration — point to deliberate data collection, not sloppy hygiene.
Why this call
Any one of these alone would be CAUTION. Together they contradict the repo's stated purpose, and two of them fire on install, which is the DO NOT CLONE line: you shouldn't have it on a machine where a stray pip install -e . could trigger it.
I read setup.py, requirements.txt, and all 4 files under src/ — roughly 95% of the executable surface. The static sweep flagged 2 install hooks, 1 exfil URL, and 0 obfuscation hits; each was opened and judged in context.
Next steps
- Report the repo to GitHub (Report content → malware) and the PyPI packages via PyPI's "Report project as malware" link.
- Warn the Discord server where you found the link.
- For async scraping, use an established library such as
httpx+selectolax, orscrapy— I can vet whichever you pick.
Connects securely to your tools. The creator never sees your data.
About this skill
Developers often clone and run code from Reddit, Discord, or GitHub without verification, risking their local machines. This skill provides a systematic framework for auditing untrusted repositories within a disposable sandbox before any code reaches your hardware.
What it does
- Metadata recon analyzes repository age, contributor history, and fork status to detect lookalike forks.
- Registry verification checks PyPI or NPM to ensure the install command targets the intended project and version.
- Static sweep scans for obfuscation, install hooks, exfil URLs, and credential probes without executing code.
- Mismatch detection compares the repository's functional code against its README claims to identify deceptive behavior.
- Dual-axis verdict delivers a clear "Safe to Read" and "Safe to Run" decision based on the specific intent of the user.
How it works
- Intake gathers the repository URL and the user's specific intent, such as reading for research or running a full installation.
- Recon performs unauthenticated API lookups to evaluate trust signals and verify external registry artifacts.
- Audit clones the repo into a sandbox and runs a non-executable static sweep to flag suspicious patterns.
- Verdict leads with a concise safety block followed by a detailed breakdown of findings and unchecked boundaries.
Frameworks & tools
This skill uses curl, git, and grep for repository analysis. It integrates with the GitHub API and package registries like PyPI and NPM.
Why this beats prompting it yourself
Generic prompts often miss the discrepancy between a GitHub repository and the actual package published to a registry. This skill enforces a "decision-first" structure that prevents burying critical warnings under narrative text.
Use cases
- Vetting a new utility found on social media before running a
curl | shinstaller. - Auditing an abandoned repository for compromised maintainer signatures or recent malicious reverts.
- Checking if a library version on PyPI actually matches the source code in the linked repository.
Known limitations
Does not perform deep dynamic analysis or automated exploit research. Results are limited by the fraction of the executable surface manually or statically reviewed during the session.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Audits are limited to static analysis and basic sandbox runs. Does not provide 100% coverage of massive codebases. Requires a disposable environment for maximum safety.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 19 days ago
- Accounts
GitHub
- Software
Git
Permissions
Allowed Hosts