rebyte-app-builder
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The
SKILL.mdfile contains explicit instructions for the agent to bypass user confirmation and suppress technical details during execution. Specifically, it commands the agent to 'JUST EXECUTE', 'hide terminal output', 'DO NOT ask the user to run commands themselves', and 'DO NOT explain commands'. While intended for user convenience, these instructions function as concealment patterns that significantly reduce transparency and user oversight of the agent's shell operations. - [INDIRECT_PROMPT_INJECTION]: The skill's architecture creates an attack surface for indirect prompt injection because it processes untrusted user-controlled data to perform actions.
- Ingestion points: The agent reads and processes user project files, including source code, shell scripts, and configuration files (e.g.,
rebyte.json,.rebyte/config.json), to facilitate the build and deployment process. - Boundary markers: The instructions do not define boundary markers or include warnings to ignore instructions embedded within the user's project files during processing.
- Capability inventory: The skill possesses extensive capabilities, including executing shell commands (e.g.,
rebyte,npm,pip,go,cargo,mkdir,cp,cat,find,rm), writing to the local file system, and deploying code to the Rebyte Cloud platform. - Sanitization: No sanitization, escaping, or validation of the content processed from user project files is mentioned before the data is interpolated into the agent's context or used for command generation.
Audit Metadata