The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text inputs provided by the user, such as snippets from emails, chats, or job descriptions, to extract and verify contractual terms.
Ingestion points: User-provided text about job offers is processed through SKILL.md instructions and passed to the scripts/check_sidework_terms.py script via a JSON input.
Boundary markers: The instructions explicitly command the agent to distinguish between different sources (e.g., written vs. verbal) and warn against assuming that chat messages constitute formal "written" terms.
Capability inventory: The skill is limited to local Python execution for data analysis; it has no capabilities for network access, file system writes, or automated tool invocation beyond standard output.
Sanitization: The bundled scripts/_common.py includes predefined regex patterns (EMAIL_PATTERN, PHONE_PATTERN, MYNUMBER_PATTERN) to identify and protect sensitive personal information within the analyzed text.
[SAFE]: No malicious patterns, such as obfuscation, hidden URLs, remote code execution, or privilege escalation, were detected. The skill's external dependencies are limited to standard Python libraries, and all logic is contained within provided scripts that follow best practices for data processing.

sidework-terms-check