The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided text to rewrite tweets, build voice profiles, and generate image prompts, creating a surface where malicious instructions embedded in the input could influence agent behavior or tool usage. 1. Ingestion points: User input provided to the main x-humanizer command (SKILL.md), tweet drafts processed for auditing (sub-skills/post-audit.md), and pasted samples used for voice profiling (sub-skills/voice-profile.md). 2. Boundary markers: None identified. The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing untrusted content. 3. Capability inventory: The skill has the capability to write to the local file system (../../references/voice-profile.md) and invoke external image generation tools (lib.illustrate and lib.refine in sub-skills/illustration.md). 4. Sanitization: No explicit sanitization, filtering, or validation of the input text is described before it is interpolated into prompts or used by tools.

x-humanizer