The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands using the GitHub CLI (gh) and npx to perform searches, preview skills, and view repository metadata. Evidence includes commands like gh skill search, gh search repos, and gh repo view found in SKILL.md.
[REMOTE_CODE_EXECUTION]: The skill uses npx skills find as a fallback method to discover new skills. This command fetches and executes the skills package from the npm registry at runtime without version verification, which could allow for the execution of arbitrary code if the package is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically GitHub repository search results and README files, creating a surface for injection attacks.
Ingestion points: Data enters the agent's context through search results and repository inspection commands in SKILL.md.
Boundary markers: The instructions lack explicit delimiters to separate external content from the agent's logic, relying on the agent to correctly parse and report on the findings.
Capability inventory: The skill allows for command execution and recommends the evaluation of third-party code, which could be exploited if an attacker-controlled repository is processed.
Sanitization: There is no evidence of input validation or content sanitization for the data retrieved from GitHub.

reuse-first