The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs the grep utility via the Bash tool to extract LaTeX symbols for notation consistency checks. This command execution is restricted to static pattern matching on local text files.
[PROMPT_INJECTION]: The skill processes external documents (LaTeX files and literature logs), which creates a surface for indirect prompt injection. Maliciously crafted content within these files could attempt to influence the audit's findings or reasoning process.
Ingestion points: workspace/paper/main.tex and workspace/logs/literature_kb.json (SKILL.md)
Boundary markers: Absent; documents are processed without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: Read/write file access, directory globbing, and grep execution via Bash (SKILL.md)
Sanitization: No evidence of input validation or content sanitization was found in the skill logic.
[SAFE]: The skill contains no hardcoded credentials, performs no external network operations, and does not use obfuscation. Its functionality is consistent with its stated purpose of mathematical verification.

theory-audit