The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted user-provided resumes and integrate this with a large local knowledge base. This creates a standard surface for indirect prompt injection.
Ingestion points: The SKILL.md file defines the user's resume (pasted text or attached file) as the mandatory primary input.
Boundary markers: Absent. The instructions do not define specific delimiters (e.g., XML tags or block markers) to separate user resume content from system instructions.
Capability inventory: The agent is restricted to using Grep and Read tools on the local knowledge/ directory. No network-write or shell execution capabilities were observed in the provided instructions.
Sanitization: The skill includes a proactive privacy instruction to 'Ignore phone number / email / ID card etc. privacy fields' during processing, which reduces the risk of sensitive data being mirrored in the output.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or sensitive local paths (like SSH keys) were found. The skill references codefun2000.com and Bilibili URLs which are consistent with the author's stated purpose and resources.

hw-interview