The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructional steering to adopt a specific analytical persona ('Extreme Rationality', 'Direct and Sharp'). These instructions guide the AI's reasoning process and output format without attempting to bypass safety filters, disclose system prompts, or override core safety guidelines.
[DATA_EXFILTRATION]: No network operations (such as curl, fetch, or wget) or sensitive file access patterns were found. The skill operates entirely within the text-based context provided by the user and its own internal reference files.
[OBFUSCATION]: The content is provided in plain text. A scan for Base64 encoding, zero-width characters, homoglyphs, and hidden Unicode tags yielded no findings. The text is clear and readable.
[REMOTE_CODE_EXECUTION]: The skill does not define any external dependencies, package installations, or remote script execution triggers. It explicitly states in SKILL.md that scripts and assets are not needed.
[DYNAMIC_CONTEXT_INJECTION]: There are no instances of the shell execution syntax (!command``) within the markdown instructions or reference files.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied decision problems, it has no dangerous capabilities (like file writing or network access) that could be exploited by malicious input. The output is limited to structured text analysis.

mungers-lattice