The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided technical text for rewriting, which presents a standard attack surface for indirect prompt injection where malicious instructions could be embedded in the input text.
Ingestion points: The skill ingests untrusted text provided by the user to be rewritten or checked against ASD-STE100 rules as described in SKILL.md.
Boundary markers: The skill includes an 'Untouchables' section in SKILL.md that explicitly instructs the agent to leave code blocks, identifiers, CLI commands, and quoted errors exact, effectively creating a boundary for technical content.
Capability inventory: The skill is restricted to text transformation and does not utilize tools for file system modification, network operations, or subprocess execution.
Sanitization: The instructions mandate that code and identifiers are not modified, which helps prevent the agent from accidentally executing or malforming embedded commands during the rewriting process.
[SAFE]: No evidence of malicious prompt injection, data exfiltration, obfuscation, or unauthorized command execution was found. The skill uses standard instructional language to guide the AI's writing style. External references to asd-ste100.org are legitimate pointers to the official standard documentation.

simple-english