The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute system commands, specifically mentioning the use of the gh (GitHub CLI) for managing issues, pnpm for build processes, and mcpc for probing tool behavior.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources including GitHub issue descriptions and web-based documentation (via WebFetch). This content is used to generate implementation plans and specifications.
Ingestion points: External content is pulled from GitHub issues using gh issue list and from documentation URLs via WebFetch.
Capability inventory: The skill has access to file system tools (Read, Glob), network tools (WebFetch), and shell execution (Bash).
Boundary markers: The instructions do not define explicit delimiters or instructions to ignore malicious patterns embedded in the external data.
Sanitization: There is no evidence of sanitization or validation of the external content before it is processed by the agent.
[DATA_EXFILTRATION]: The skill facilitates reading local source code and project internal data, which, when combined with its ability to create GitHub issues and perform web requests, creates a theoretical path for exfiltrating code or configuration if the agent is manipulated.

dig