claude-api

Pass

Audited by Gen Agent Trust Hub on Mar 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill provides implementation patterns for autonomous agents (e.g., Coding Agent, Support Agent) that process data from external sources such as project requirement files, user messages, and tool outputs.
  • Ingestion points: Data from features.md, user-provided support queries, and financial market data tool results.
  • Boundary markers: The architectural examples provided do not specify the use of delimiters or explicit instructions for the agent to disregard potential instructions embedded in the ingested data.
  • Capability inventory: The agents are designed to use various tools including file editing, command-line execution (Bash), and web interaction (WebFetch).
  • Sanitization: The provided code snippets demonstrate functional logic but do not include explicit input sanitization or validation steps for content retrieved from external sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 22, 2026, 04:49 PM