openai-evals-tuning
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements prompt templating for evaluation runs, which creates a surface where malicious instructions within input data could influence model behavior.
- Ingestion points: External JSONL data is ingested and interpolated into prompts in
samples/eval-create-and-run.mdandreferences/evals/working-with-evals.md. - Boundary markers: The prompt templates demonstrated in
samples/eval-create-and-run.mddo not use protective delimiters or specific instructions to ignore embedded commands in the input data. - Capability inventory: The skill provides functionality for network communication with the OpenAI API via scripts (
scripts/evals.md,scripts/fine-tuning-jobs.md) and Python samples (samples/eval-create-and-run.md). - Sanitization: No validation or sanitization of the input data is performed before it is interpolated into the evaluation prompts.
Audit Metadata