The Agent Skills Directory

[SAFE]: The instructions describe legitimate usage of a skill management CLI tool. No dangerous commands, privilege escalation, or persistence mechanisms were found.
[INDIRECT_PROMPT_INJECTION]: The skill utilizes the output of environment-scanning commands like cortex skills recommend and cortex status to drive agent actions. While this constitutes a data ingestion surface where external state (such as file names in a git diff) could theoretically influence agent behavior, it is the intended functional purpose of the tool and is managed within the CLI's expected operating context. Ingestion points: Command output from cortex skills recommend and cortex status referenced in SKILL.md. Boundary markers: None provided in the instruction set. Capability inventory: Execution of specific cortex CLI commands for status, feedback, and rating. Sanitization: No explicit sanitization or input validation logic described in the instructions.

cortex-skills-loop