model-routing-preset-builder
Warn
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to run the
npx skillscommand to install and verify policy changes on the host system. - [REMOTE_CODE_EXECUTION]: By invoking
npx skills, the skill triggers the download and execution of code from the public npm registry at runtime. - [EXTERNAL_DOWNLOADS]: The use of
npxinvolves fetching external packages from the npm registry, which is a potential source of unverified code. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection via processed data.
- Ingestion points: The skill reads operator-supplied changes and local configuration baselines (e.g.,
model-capability-router,references/gpt61-sol-20260929.md). - Boundary markers: There are no explicit markers or instructions to delimit or ignore potentially malicious instructions embedded in the ingested operator requests or reference files.
- Capability inventory: The skill has the capability to execute shell commands via
npx skillsand write changes to a source repository. - Sanitization: The instructions do not specify any sanitization, validation, or filtering of the operator-supplied data before it influences the command execution.
Audit Metadata