The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install an external Python package gateguard-ai from a public registry. This package does not originate from a known trusted organization or the verified vendor infrastructure.
[COMMAND_EXECUTION]: The skill requires the execution of gateguard init and references a local JavaScript hook at scripts/hooks/gateguard-fact-force.js which was not provided for analysis.
[REMOTE_CODE_EXECUTION]: The combination of installing an unverified external package and executing its initialization CLI tool constitutes a remote code execution risk.
[INDIRECT_PROMPT_INJECTION]: The gatekeeper logic forces the agent to read and process untrusted external data (file imports, data schemas) and verbatim user instructions. This content is then used to decide whether to allow or block actions.
Ingestion points: Reads source code for imports via Grep/Bash and reads data structures to identify schemas; quotes user instructions.
Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" warnings for the data being analyzed during the fact-finding stage.
Capability inventory: The skill utilizes Edit (file-write), Bash (subprocess/command execution), Grep, and Glob tools.
Sanitization: There is no technical enforcement of data sanitization; the skill relies on the model to manually redact values.

gateguard