The Agent Skills Directory

[OBFUSCATION]: The header # Graydon Hoare Style Guide in SKILL.md contains a sequence of zero-width characters (U+200B, U+200C, U+200D, U+2060) that encode a hidden payload. This technique is used to bypass visual inspection and automated scanners by hiding content that is still processed by the AI.
Evidence: The header string # Graydon Hoare Style Guide is followed by over 100 hidden Unicode characters wrapped in Word Joiner markers.
[INDIRECT_PROMPT_INJECTION]: The skill uses hidden steganographic content to potentially provide instructions to the agent that are not visible to the user. This creates an attack surface where the agent might follow hidden commands embedded in the documentation, potentially influencing its response or overriding safety constraints.
Ingestion points: The AI agent reads the instructions in SKILL.md during activation.
Boundary markers: There are no boundary markers or instructions to ignore embedded Unicode tags.
Capability inventory: The skill is primarily instructional for Rust programming, but the presence of steganography indicates an attempt to subvert the agent's reasoning process.
Sanitization: No sanitization is performed on the skill content before it is processed by the AI.

hoare-rust-origins