The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user input (prompts to be optimized) and output a modified version.
Ingestion points: User messages identified in SKILL.md (Step 0).
Boundary markers: No explicit delimiters or 'ignore' instructions are used for the input, though strict output formatting is enforced.
Capability inventory: The skill performs only natural language processing and lacks capabilities for subprocess execution, network access, or file system modifications.
Sanitization: No specific sanitization or escaping is performed, as the skill aims to preserve the literal meaning and factual data of the original input.
[SAFE]: The skill does not perform any network operations, access sensitive files, or include hardcoded credentials. It explicitly instructs the agent to refrain from executing the prompt being optimized unless specifically told otherwise, which serves as a safety boundary.

optimiza-prompt