practice-cycle

Pass

Audited by Gen Agent Trust Hub on Jul 17, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill processes 'observed results' from task execution, creating a potential surface for indirect prompt injection where untrusted data could attempt to misdirect the agent's logic.\n
  • Ingestion points: SKILL.md (Start template for observed results).\n
  • Boundary markers: The skill relies on natural language instructions ('Separate observation from interpretation') rather than technical delimiters or sandboxing for external content.\n
  • Capability inventory: No subprocess calls, exec/eval, file-write, or network operations are defined in SKILL.md, agents/openai.yaml, or the referenced files.\n
  • Sanitization: No explicit sanitization or validation of the ingested task data is performed.\n- [EXTERNAL_DOWNLOADS]: The file references/sequence.md refers to the author's official GitHub repository (github.com/lidessen/skills.git) as the canonical source for the philosophical principles used in the reasoning loop. This is a vendor-owned resource.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 17, 2026, 06:32 AM
Security Audit — agent-trust-hub — practice-cycle