jd-interview-prep

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture relies on ingesting untrusted external data, which creates a potential surface for indirect prompt injection attacks.
  • Ingestion points: The workflow processes user-provided Job Descriptions and Resumes (SKILL.md, PROMPT.md) and retrieves data from third-party websites during the company research phase (SKILL.md, Step 2).
  • Boundary markers: The skill uses specific delimiters like ---JD START--- in its prompts to separate user content, but it does not provide consistent instructions for the agent to treat all ingested data as data rather than potential instructions.
  • Capability inventory: The skill utilizes the agent's ability to read local files (e.g., candidate-context.md), perform web searches using available tools, and write various markdown output files to the workspace.
  • Sanitization: The instructions lack specific guidance for the agent to sanitize or escape external content before interpolating it into the final reports or using it in subsequent decision-making steps, which could allow maliciously crafted input to influence agent output.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:34 PM
Security Audit — agent-trust-hub — jd-interview-prep