verify

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it performs AI-driven analysis (Claude analysis) on untrusted source code and finding metadata provided in the verification context.\n
  • Ingestion points: The workflow ingests data from .appsec/findings.json and the source code at the location of the reported vulnerability.\n
  • Boundary markers: There are no explicit delimiters or instructions provided to isolate the code being analyzed from the agent's operational instructions, increasing the risk that embedded comments could be interpreted as commands.\n
  • Capability inventory: The skill uses file-read tools to load code, file-write tools to update .appsec/findings.json and .appsec/fixed-history.json, and has the capability to execute system scanners via CLI.\n
  • Sanitization: No mechanisms for sanitizing or filtering untrusted code content are described before the AI evaluates the fix integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 06:45 PM