The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes external task data and has access to powerful system tools.
Ingestion points: The <task> argument defined in the frontmatter and the data processed during the delegated execution phase.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the task data.
Capability inventory: The skill is permitted to use Bash, Write, Edit, and Task tools, providing a path for command execution if task data contains malicious instructions.
Sanitization: There is no evidence of escaping, validation, or filtering of the external task content before it is processed by the agent's workflow.
[SAFE]: The skill's primary intent is to increase safety and quality. It incorporates a 'Pressure Resistance' framework that explicitly instructs the agent to never skip security-sensitive steps, which serves as a safeguard against user-induced policy bypass.

with-anti-rationalization