# I statically scanned 618 public agent skills. Here's what the scanner actually found

> Source: <https://dev.to/vyankatesh/i-statically-scanned-618-public-agent-skills-heres-what-the-scanner-actually-found-2ppc>
> Published: 2026-11-20 12:09:47+00:00

**Recommendation:** Treat public agent skills as a supply-chain surface: read what skills with bundled scripts do with your credentials before enabling them, and triage static findings by capability rather than severity label.
More in the playbook "Securing agents and AI tools": <https://wpnews.pro/playbooks/agent-security>

**Disclosure:** I ran this study with an AI assistant (Claude), which wrote the scripts, did the analysis and drafted the paper. The assistant also did the manual spot-check described below, so that part is **not** an independent human review. Everything here is aggregate: no skill, author or repo is named.

Agent skills are folders of instructions (a `SKILL.md`) plus optional scripts that coding agents load on demand. The text is read by the model as trusted guidance, and the scripts can run with your permissions. That makes public skill collections a supply-chain surface. So I asked a simple question: **how risky do public skills look to a static scanner?**

`agent-skills`, `claude-skills`, `claude-code-skills`), then drew a seeded, stratified sample across four star ranges (0-9, 10-99, 100-999, 1000+).`--no-llm`). 618 scanned fine; 4 timed out and were excluded.
The most common categories were "MCP least privilege" (27% of skills), "excessive agency" (25%), privilege escalation (18%) and data exfiltration (16%). Prompt-injection findings appeared in 6.5% of skills.

Of the 25 skills with a CRITICAL finding, 23 were CRITICAL only because of "data flow" findings, usually an environment variable (like an API key) flowing into an HTTP request. In the ones I read, the destination was a fixed first-party API endpoint. The other two were outdated dependencies pinned in lockfiles.

I also read 10 randomly chosen skills. Of the 3 flagged HIGH or CRITICAL, **none** warranted that severity in my judgment. One "memory poisoning" hit was a mocking library's docs saying "clear call history". That's 3 skills judged by one non-human reviewer, so treat it as a hint that severity labels over-state risk, not as an error rate.

I did not find a skill I'd call intentionally malicious. That is **not** a claim that none exist: static scanning can't see runtime behaviour, and I didn't estimate how many real problems it misses.

SkillSpector is a static scanner: it lowers risk but doesn't remove it.

Use static scores to **triage capabilities** (network, credentials, subprocess, unpinned dependencies), not to decide who's malicious. Be most careful with skills that ship scripts, and read what a skill does with your credentials before you enable it. Larger, behaviourally validated studies are the right way to find out how common genuinely malicious skills are.

Related work I found useful (all on arXiv): [Agent Skills in the Wild](https://arxiv.org/abs/2601.10338), ["Do Not Mention This to the User"](https://arxiv.org/abs/2602.06547), [Exploring the Emerging Threats of the Agent Skill Ecosystem](https://arxiv.org/abs/2605.28588), [SkillJect](https://arxiv.org/abs/2602.14211).

*Preprint-style write-up, not peer reviewed.*

*By Vyankatesh Rohokale*
