skills/vinta/hal-9000/best-practices/Gen Agent Trust Hub

best-practices

Pass

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because its primary function involves fetching and processing data from external, untrusted sources like documentation websites and search engines.
  • Ingestion points: Untrusted data enters the agent context via 'WebSearch', 'find-docs' (Context7), and any URLs supplied by the user.
  • Boundary markers: XML-style tags such as and are used in the prompt templates to provide structure and help separate retrieved data from agent instructions.
  • Capability inventory: The skill is configured with access to 'Bash' for shell execution and network tools, which represent capabilities that could be exploited if malicious content were processed without proper safeguards.
  • Sanitization: The skill's instructions do not include specific steps for sanitizing or validating the retrieved external content before synthesis. This risk is inherent to the research-oriented nature of the skill.
  • [SAFE]: No evidence of malicious behavior, obfuscation, hardcoded credentials, or unauthorized system access was identified. The skill's use of 'ctx7' is consistent with its stated purpose of documentation retrieval.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 22, 2026, 06:08 AM