data-processing
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted external data from configuration files.
- Ingestion points: Processes files such as
package.json,.env,docker-compose.yml, and Kubernetes manifests. - Boundary markers: Absent. The skill provides direct commands to read these files without delimiters or warnings to ignore embedded instructions.
- Capability inventory: Utilizes shell execution via the
Bash Readtool to runjqandyqcommands, including in-place file modifications (yq -i). - Sanitization: None. The skill does not validate or escape the content of the data files before processing.
- [Command Execution] (SAFE): Use of
jqandyqis standard practice for structured data manipulation. The patterns provided are idiomatic and follow common utility usage. - [Data Exposure] (LOW): The skill description explicitly mentions triggering on
.envfiles. While no exfiltration logic is detected, processing environment files can inadvertently expose sensitive credentials to the agent's context.
Audit Metadata