The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions in README.md use a curl | sh pattern to download and execute a script from skil.sh. This pattern is high-risk as it executes remote code from a source outside of the trusted vendor list.
[COMMAND_EXECUTION]: The workflow in SKILL.md interpolates variables like {number} and keyword directly into shell commands for the gh and git tools. If these variables are derived from malicious GitHub issue content, they could lead to command injection.
[EXTERNAL_DOWNLOADS]: The skill fetches an external script from the skil.sh domain, which is not recognized as a trusted organization or service.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub issues to drive its analysis and response generation. * Ingestion points: The gh issue view command fetches issue titles, bodies, and comments. * Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the fetched content. * Capability inventory: The agent has access to gh, git, and grep. * Sanitization: No sanitization process is defined for the external issue content before analysis.

fiftyone-issue-triage