The Agent Skills Directory

[SAFE]: No malicious behavior, unauthorized network requests, or credential harvesting patterns were detected in the skill code or metadata.- [PROMPT_INJECTION]: The skill uses assertive language in its description and instructions (e.g., 'Must always apply', 'Portability (required)') to define its writing style and execution protocols. These are functional instructions for the agent's behavior and do not attempt to bypass safety constraints.- [INDIRECT_PROMPT_INJECTION]: The skill's core purpose is to process and transform external text, which serves as an ingestion point for untrusted data. * Ingestion points: The skill processes text provided for 'unsloping' as defined in SKILL.md. * Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present. * Capability inventory: While SKILL.md and references/runtime.md mention agent orchestration capabilities like 'parallel' and 'model_role', these are not linked to the content being processed. * Sanitization: No sanitization of the processed text is described. * Analysis: The risk is negligible as the skill lacks execution sinks (like file writing or command execution) that could be triggered by instructions embedded in the text being rewritten.

unslop