read-repo-references

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: HIGHPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • PROMPT_INJECTION (HIGH): The skill is designed to ingest and follow instructions or patterns from the .references/ directory and external URLs. An attacker can place malicious instructions in these locations to hijack the agent's behavior during the 'Apply learnings' phase.
  • Ingestion points: .references//REF.md and external URLs
  • Boundary markers: None present
  • Capability inventory: cat, ls, WebFetch/Browsing
  • Sanitization: None
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill directs the agent to fetch content from arbitrary URLs defined in metadata, which could lead to the ingestion of malicious scripts or data into the agent's context.
  • COMMAND_EXECUTION (LOW): Uses standard shell commands (ls, cat) to interact with the local filesystem to retrieve reference data.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 17, 2026, 12:23 AM