The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied job offer details which could contain malicious instructions. However, the risk is mitigated because the skill is strictly restricted to formatting and calculation. It lacks capabilities such as network access, file system modifications, or shell execution that could be exploited by an injection attack.
Ingestion points: Job offer data is ingested from a user-provided JSON file processed by scripts/compare_offers.py.
Boundary markers: The agent is explicitly instructed in SKILL.md to adhere to fixed report templates and not to provide recommendations or autonomous responses.
Capability inventory: The script only uses standard Python libraries for JSON parsing and decimal math. No network or system-level capabilities are requested or used.
Sanitization: The Python script implements type and structure validation for the incoming data, ensuring only expected fields are processed.
[SAFE]: No security issues were detected. The skill follows security best practices, including validating input data types and explicitly restricting the AI agent from taking autonomous actions or making subjective recommendations. The provided script uses standard libraries and does not engage in network or sensitive file operations.

offer-comparison