openserv-agent-sdk

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is authored by the platform vendor (openserv-labs) and utilizes official, scoped packages and trusted infrastructure. It follows security best practices by using environment variables for credentials and providing clear error handling in examples.\n- [PROMPT_INJECTION]: The skill examples demonstrate the ingestion of untrusted data from user-defined tasks and its interpolation into prompts for LLM processing. This represents a potential surface for indirect prompt injection, which is handled as a standard risk factor for agentic systems.\n
  • Ingestion points: The 'run' function in examples such as 'examples/capability-example.ts' accepts data through 'args' (e.g., 'args.text').\n
  • Boundary markers: Examples use direct string interpolation without specific delimiters or guardrails to prevent command injection within the data.\n
  • Capability inventory: Templates provide access to methods for generating LLM content, creating tasks, and managing files.\n
  • Sanitization: No explicit data sanitization or instruction-guarding is implemented in the template code, as is common for educational SDK examples. Developers are encouraged to implement these measures for production use.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 07:54 AM