diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external files and network artifacts to drive debugging actions.\n
- Ingestion points:
CONTEXT.md, Architecture Decision Records (ADRs), and user-provided inputs captured byscripts/hitl-loop.template.sh.\n - Boundary markers: Absent. No explicit instructions are provided to use delimiters or ignore embedded instructions when reading these files.\n
- Capability inventory: Broad shell access for executing tests,
curlcommands, CLI tools, and headless browsers as outlined inSKILL.md.\n - Sanitization: Absent. No filtering or escaping is specified for data read from files or user inputs before processing.\n- [DYNAMIC_EXECUTION]: The skill relies on the agent generating and executing code at runtime to build reproduction loops.\n
- Evidence: Phase 1 of
SKILL.mddirects the agent to construct and run various diagnostic tools, such as HTTP scripts, CLI invocations, and throwaway harnesses.
Audit Metadata