black-tortoise-agent-orchestration
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The skill ingest data from local documentation files (docs/INDEX.md, AGENTS.md) and external library documentation via the context7 MCP tool. This creates a surface for indirect prompt injection where an attacker could influence the agent by placing malicious instructions in those data sources. • Ingestion points: docs/INDEX.md, AGENTS.md, README.md, and library docs retrieved via context7. • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the skill definition. • Capability inventory: The skill can read local files, call various MCP tools, and spawn subagents using agent/runSubagent. • Sanitization: No sanitization or validation of the ingested content is defined.
- [Command Execution] (SAFE): The skill references local development commands such as pnpm run lint and pnpm run architecture:gate. However, it explicitly states these should not be executed automatically (不自動執行), reducing the risk of unauthorized execution.
- [Data Exposure] (SAFE): While the skill reads project-level documentation, it does not target sensitive system files (e.g., SSH keys, credentials) or exhibit patterns for data exfiltration.
Audit Metadata