The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill defines local execution of Python scripts for content analysis. The provided script brand_voice_analyzer.py is safe, using only standard libraries (re, json, sys) to compute readability scores and keyword frequencies. It does not perform any dangerous system calls, file modifications, or network operations.\n- [DATA_EXPOSURE] (SAFE): While the tools are designed to read user-provided text files, there is no evidence of hardcoded credentials, sensitive path access, or unauthorized data transmission. The script outputs metadata (counts and scores) rather than echoing raw input, which minimizes unintended data exposure to the agent context.\n- [INDIRECT_PROMPT_INJECTION] (SAFE): The skill has a data ingestion surface (analyzing external files), but the script's output is restricted to statistical metrics and static recommendations. This architecture prevents instructions embedded in analyzed content from being interpreted as commands by the agent.\n- [METADATA_POISONING] (SAFE): Metadata fields and reference guides contain standard marketing documentation and do not include any hidden instructions or obfuscated triggers.

content-creator