The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the gateguard-ai Python package via pip. This is an external dependency provided by a community author and is not associated with a trusted vendor or well-known service.
[COMMAND_EXECUTION]: The documentation requires running the command gateguard init to generate a local configuration file (.gateguard.yml) and initialize the tool's environment.
[INDIRECT_PROMPT_INJECTION]: The skill's logic requires the agent to "Quote the user's current instruction verbatim" as part of its multi-stage gate process (Edit, Write, and Bash gates).
Ingestion points: The gate prompts in SKILL.md process live user instructions as part of the mandatory investigation criteria.
Boundary markers: Absent; the instructions do not define delimiters (like triple backticks or specific tags) or provide "ignore embedded instructions" warnings for the repeated user content.
Capability inventory: The skill controls access to file editing, file creation, and destructive Bash commands (including rm -rf, git reset --hard, and drop table).
Sanitization: While the skill recommends using redacted or synthetic values when investigating data files, it lacks sanitization or validation mechanisms for the user instructions it requires the agent to quote.

gateguard