The Agent Skills Directory

[SAFE]: The skill is implemented as a set of markdown instructions and a JSON schema without any accompanying scripts or binary files, minimizing the execution risk.
[SAFE]: No network access, file system interaction, or credential management is defined. The skill is self-contained and limited to processing text inputs.
[PROMPT_INJECTION]: The skill ingests untrusted text via the resume, jd, and sourceTexts fields, creating a potential surface for indirect prompt injection. However, since the skill has no access to external tools or system resources, the risk is negligible. The instructions include a specific 'truthfulness' assessment to detect and flag fabricated content within the inputs.
Ingestion points: resume, jd, and sourceTexts parameters in the skill input schema.
Boundary markers: The instructions do not define specific delimiters to separate untrusted user data from the agent's instructions.
Capability inventory: The skill is restricted to diagnostic analysis and structured JSON output with no access to dangerous tools.
Sanitization: No input filtering or sanitization logic is present in the skill definition.

review