The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (HIGH): The skill suggests downloading a binary directly from an untrusted GitHub repository and executing it. \n
Evidence: curl -sL -o ls-lint https://github.com/loeffel-io/ls-lint/releases/download/v2.3.1/ls-lint-linux-amd64 followed by chmod +x ls-lint. \n
Risk: The organization loeffel-io is not on the trusted sources list. Downloading and executing unverified binaries is a significant security risk as it bypasses package registry safety checks. \n- [Privilege Escalation] (MEDIUM): The skill instructs the agent to grant execution permissions to a downloaded file using chmod +x. \n
Evidence: chmod +x ls-lint in the installation section. \n- [Indirect Prompt Injection] (LOW): The skill processes untrusted data from the local filesystem (file and directory names) which could potentially be crafted to influence agent behavior or exploit vulnerabilities in the linter binary. \n
Ingestion points: Filesystem paths and names processed by ls-lint (SKILL.md). \n
Boundary markers: Absent. \n
Capability inventory: Execution of ls-lint binary and npx commands which read local filesystem data. \n
Sanitization: None specified.

ls-lint