diagnose-software-failures
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary function is to guide an agent through a structured debugging workflow. The instructions promote evidence-based diagnosis rather than speculative patching, which is a sound engineering practice. No findings related to credentials, malicious commands, or unauthorized exfiltration were detected.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of indirect prompt injection from processing external logs and issue reports. It contains specific defensive instructions in
failure-investigation-organizational-controls-and-scale.mdandfailure-investigation-concepts-contracts-and-invariants.mdto treat external inputs as untrusted and to maintain authorization boundaries, which serves as a mitigation for this attack vector. - [EXTERNAL_DOWNLOADS]: The skill references several external resources for debugging documentation (e.g., Google SRE, Clang Sanitizers, LLVM, OpenAI/Anthropic documentation, and IETF RFCs). These are all reputable and well-known industry sources and do not involve the execution of untrusted remote code.
Audit Metadata