diagnosing-bugs

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill facilitates the creation and execution of dynamic code, such as reproduction scripts and throwaway harnesses, which are necessary for identifying and fixing software regressions.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources like system logs, network request traces, and code files to inform its diagnosis process.\n
  • Ingestion points: Log files, network traces (HAR), and artifacts provided by the user (SKILL.md).\n
  • Boundary markers: The skill relies on structured phases rather than explicit prompt boundaries for ingested data.\n
  • Capability inventory: The skill uses shell command execution, file system interaction, and network tools (SKILL.md).\n
  • Sanitization: External artifacts are processed for analysis without explicit sanitization steps.\n- [COMMAND_EXECUTION]: The skill includes a bash template (scripts/hitl-loop.template.sh) for user-assisted debugging. The script follows secure scripting practices, including the use of read -r and controlled variable assignment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 10:56 AM
Security Audit — agent-trust-hub — diagnosing-bugs