The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub issue comments which can contain malicious instructions to influence agent behavior.
Ingestion points: Retrieves issue text and all comments via gh issue view commands in SKILL.md.
Boundary markers: Lacks explicit delimiters or instructions to treat issue content as data rather than instructions.
Capability inventory: Includes the ability to execute gh CLI commands.
Sanitization: No filtering or sanitization of the fetched GitHub content is performed before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to run gh CLI commands using direct interpolation of user input.
Evidence: The pattern gh issue view $ARGUMENTS --repo [REPO] is used in SKILL.md without validation of the $ARGUMENTS variable, potentially allowing command flag injection.

issue-summary