The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified in the skill's instructions, templates, or documentation.
[DATA_EXPOSURE]: The skill does not access sensitive local file paths or include hardcoded credentials. It features a dedicated 'Confidentiality Scrub' step designed to protect the user by flagging internal project names and trade secrets for generic replacement.
[REMOTE_CODE_EXECUTION]: There are no remote code execution patterns, third-party package dependencies, or dynamic code generation mechanisms. The skill relies on static templates for document generation.
[PROMPT_INJECTION]: The skill instructions include robust safety rules that prohibit the agent from offering legal advice, guaranteeing patentability, or bypassing safety filters. No adversarial injection patterns were found.
[COMMAND_EXECUTION]: No shell command execution or dynamic context injection patterns were detected in the skill definition.
[OBFUSCATION]: No obfuscated content, hidden URLs, or multi-layer encoding were found. The binary .skill file is a standard ZIP archive containing the provided skill source files.

patentforge-provisional