mtthwmllr/skill-safety-auditor
Audits a Claude Code skill for security risks in three modes: before download (from a URL or install command), after download but before install (from a .skill file), or after install (from a local skills directory). Use this skill whenever a user is about to install a skill from any source — including GitHub URLs, git clone commands, npx/npm commands, curl/wget downloads, pip installs, marketplace links, or raw SKILL.md URLs. Also trigger when a user asks "is this skill safe?", "should I trust this skill?", "can you check this before I install it?", "audit this skill", or pastes any link to a skill repository or .skill file. If a user mentions installing ANY skill, proactively offer to audit it first — do not wait for them to ask.
97
97%
Does it follow best practices?
Impact
99%
1.28xAverage score across 5 eval scenarios
Advisory
Suggest reviewing before use
Security
2 findings — 2 medium severity. This skill can be installed but you should review these findings before use.
W011: Third-party content exposure detected (indirect prompt injection risk)
What this means
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
Why it was flagged
Third-party content exposure detected (high risk: 0.80). The skill's Mode 1 workflow explicitly uses WebFetch to retrieve raw SKILL.md and referenced scripts from public sources (e.g., GitHub/raw URLs) and then reads and interprets that untrusted, third‑party content as part of its audit, which could enable indirect prompt‑injection to influence decisions.
W012: Unverifiable external dependency detected (runtime URL that controls agent)
What this means
The skill fetches instructions or code from an external URL at runtime, and the fetched content directly controls the agent’s prompts or executes code. This dynamic dependency allows the external source to modify the agent’s behavior without any changes to the skill itself.
Why it was flagged
Potentially malicious external URL detected (high risk: 0.90). Yes — in Mode 1 the skill uses WebFetch to fetch SKILL.md at runtime from raw GitHub URLs (e.g. https://raw.githubusercontent.com/USER/REPO/BRANCH/SKILL.md), and the fetched SKILL.md directly controls the agent's prompts/instructions.
- Audited
- Security analysis