The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing external, potentially untrusted text data via its diagnostic scripts.
Ingestion points: The scripts scripts/humanize.py and scripts/detect.py read user-supplied text files (e.g., draft.txt) provided via command-line arguments as their primary input for analysis.
Boundary markers: The scripts do not utilize boundary markers or delimiters to separate untrusted text from script logic, nor do they provide warnings to the agent regarding potential instructions embedded within processed content.
Capability inventory: The script scripts/humanize.py possesses file-writing capabilities through the -o or --out command-line flag. Both scripts use the re module to execute potentially complex regular expression patterns on user-provided text.
Sanitization: There is no explicit sanitization, filtering, or escaping of the external text content to prevent the agent from inadvertently interpreting embedded instructions found in the drafts it is directed to analyze.

ig-human