The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script (revision_patch.py) for processing Markdown files. The script is designed with safety guards, such as the refuse_same function, which prevents the accidental overwriting of source files during the anchoring and application phases.- [PROMPT_INJECTION]: The skill processes external Markdown data which could contain malicious instructions (Indirect Prompt Injection surface). This risk is addressed through an integrity-first workflow: each block is assigned a unique ID and a SHA256 hash. The check and apply commands verify these hashes against a manifest, ensuring only intended blocks are modified and preventing silent document manipulation.- [SAFE]: Analysis of the skill instructions and associated Python scripts found no signs of obfuscation, hardcoded credentials, or persistence mechanisms. The tool operates entirely locally without external network dependencies.

rw-revision-patch