dspy-evaluation-suite
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: CRITICALPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The skill possesses a surface for indirect prompt injection through the processing of untrusted evaluation data.
- Ingestion points: The
devsetinput (list ofdspy.Exampleobjects) is processed by the agent inEvaluationSuite.evaluateandEvaluationSuite.compare. - Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the code or prompts.
- Capability inventory: The skill utilizes
dspy.Modulewhich makes network calls to LLMs and has theWritetool enabled in the YAML manifest to export JSON reports viaexport_report. - Sanitization: No sanitization or validation of the text content within
devsetexamples is performed before they are passed to theevaluator. - Security Scanners (SAFE): The automated scanner alert for
logger.info(URL:Blacklist) is a false positive. The scanner misidentified standard Python logging callslogger.info(f"Evaluating: {name}")as a malicious domain or URL string.
Recommendations
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata