The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input regarding personal emotions and life goals, which presents a surface for indirect prompt injection.
Ingestion points: User prompts triggered by terms like 'happiness', 'desire', and 'anxiety' are processed by the agent using the provided methodology in SKILL.md.
Boundary markers: The skill includes explicit boundary markers in the description and Execution steps, specifically instructing the agent to stop the process and suggest professional help if clinical depression symptoms are detected.
Capability inventory: The skill lacks any functional capabilities such as file system access, network operations, or shell command execution, which effectively neutralizes the risk of an injection attack causing system-level harm.
Sanitization: No explicit data sanitization is performed on input, but the output is restricted to conceptual advice.
[SAFE]: The skill is entirely composed of Markdown and JSON metadata. It does not include any Python or Node.js code, nor does it request access to any system tools or external APIs. All identified triggers and behaviors are consistent with its stated purpose of providing self-help advice.

happiness-skill