The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled Python script (scripts/check_trial_terms.py) to calculate trial work durations, weekly hours, and compensation rates based on user-provided JSON input.
[EXTERNAL_DOWNLOADS]: References informational and regulatory URLs including the Japanese Ministry of Health, Labour and Welfare (mhlw.go.jp) and the Otameshi Tenshoku service website (otame4.work) for verifying labor laws and service terms.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data including candidate resumes, job descriptions, and chat transcripts.
Ingestion points: Trial notes, candidate experience descriptions, and employer job listings (SKILL.md).
Boundary markers: The instructions mandate a clear distinction between private reasoning/analysis and the draft content intended for the employer (SKILL.md).
Capability inventory: Limited to local Python script execution for arithmetic and validation; no network exfiltration or persistent file system operations are present.
Sanitization: External input is processed through a common validation utility (scripts/_common.py) that enforces data types and utilizes regular expressions to identify and potentially protect sensitive personal information patterns.

otameshi-candidate