The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to run or install the skills-ref validator from a remote GitHub repository (github.com/agentskills/agentskills) using uvx and uv tool install. While these are expected for the skill's functionality, they involve executing code from a source outside the trusted vendors list.
[COMMAND_EXECUTION]: The skill includes shell scripts and documentation for executing system commands like git, jq, and bash to manage metadata and validate files. These commands are used to automate repository management tasks.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface (Category 8) when processing third-party SKILL.md files.
Ingestion points: The validate-skills-repo.sh script and the skills-ref tool read and process skill files from the local filesystem.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the maintenance scripts or documentation.
Capability inventory: The skill environment supports shell script execution, remote tool installation, and repository manipulation.
Sanitization: The provided scripts do not perform content sanitization or validation of the data within the SKILL.md files before processing.

agentskills-io