error-diagnostics-error-analysis
Pass
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Potential for indirect prompt injection through the analysis of untrusted diagnostic data.
- Ingestion points: Diagnostic context, logs, and stack traces are ingested via the
$ARGUMENTSparameter inSKILL.md. - Boundary markers: The skill does not define explicit delimiters or instructions to ignore commands embedded within the provided diagnostic data.
- Capability inventory: The skill performs high-level analysis and suggests implementation fixes, providing a mechanism for malicious input to influence subsequent system changes.
- Sanitization: Although the skill mandates the redaction of secrets and PII, it lacks logic to sanitize or escape instructions that may be embedded in the analyzed logs.
Audit Metadata