failure-oriented-instrumentation

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to analyze untrusted external failure data like stack traces and error messages.
  • Ingestion points: Failure information is ingested during the analysis phase described in SKILL.md.
  • Boundary markers: The skill lacks instructions for using delimiters or boundary markers to prevent the execution of instructions within the ingested data.
  • Capability inventory: The agent is empowered to modify source code and execute applications as part of the debugging process.
  • Sanitization: There is no mention of sanitizing or validating the failure data before processing.
  • [COMMAND_EXECUTION]: The skill utilizes powerful runtime hooks and debugging interfaces for instrumentation.
  • Evidence: Includes templates for GDB scripting in the C/C++ reference and sys.settrace in the Python reference to monitor program state.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 10:20 PM