experiment-pipeline
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill performs standard instructional tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data ("experiment doc") and uses it to generate files on the local filesystem.
- Ingestion points: The agent reads an external experiment document to generate the pipeline.
- Boundary markers: No explicit delimiters are instructed to be used when interpreting the input document.
- Capability inventory: The skill utilizes file writing capabilities to save the output to
docs/experiments-pipeline/. - Sanitization: No specific sanitization or filtering of the input content is mentioned before it is written to the filesystem.
- [COMMAND_EXECUTION]: No shell commands, subprocess spawning, or script executions are present in the instructions.
- [DATA_EXFILTRATION]: The skill does not perform any network operations or access sensitive credentials.
Audit Metadata