The Agent Skills Directory

[PROMPT_INJECTION]: The skill directly interpolates the $ARGUMENTS variable into the instructions for the advisor agent. Without sanitization, a user could provide malicious input intended to override the agent's defined behavior or safety constraints.
[DYNAMIC_EXECUTION]: The skill implements a feature to forward other skills by parsing a name from the user-provided arguments (e.g., /unslop). The advisor agent is then instructed to load and execute this dynamically specified skill using its tools, which could lead to the execution of unintended or malicious skills if the input is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where untrusted user input significantly influences the context and capabilities of a secondary agent.
Ingestion points: The $ARGUMENTS placeholder in SKILL.md which receives user input at runtime.
Boundary markers: The input is presented within a 'User's request' section, but there are no explicit delimiters or instructions to the advisor agent to treat this content as purely data or to ignore embedded commands.
Capability inventory: The advisor agent is granted the ability to read files by path, create agents, and load/execute other skills via the Skill tool.
Sanitization: The skill does not perform validation or sanitization on the user-provided request or the parsed skill name before it is used to drive agent behavior.

paseo-advisor