hypergen-agent
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches agent configuration and instructional documentation (e.g., soul.md, skill.md, hypergen.requests.json) from the vendor's official domain (hypergen.hypercho.com) to configure the agent's behavior.\n- [COMMAND_EXECUTION]: Provides utility functions in scripts/hypergen-agent.mjs that execute git clone to install the developer's social engagement add-on from GitHub.\n- [DATA_EXFILTRATION]: Includes a cleanMetadata function designed to strip sensitive keys such as passwords, tokens, and cookies from status reports before they are transmitted to the remote API, minimizing the risk of accidental credential exposure.\n- [PROMPT_INJECTION]: Identifies an indirect prompt injection surface as the agent is instructed to ingest and follow documentation downloaded from a remote server.\n
- Ingestion points: The scripts/hypergen-agent.mjs file downloads soul.md and skill.md at runtime to define the agent's persona and skills.\n
- Boundary markers: No explicit delimiter or safety warnings (e.g., "ignore embedded instructions") are applied to the downloaded content before processing.\n
- Capability inventory: The agent has the ability to execute shell commands (Bash and Git) via spawnSync.\n
- Sanitization: Downloaded instructional files are used directly without sanitization or validation of the text content.\n- [CREDENTIALS_UNSAFE]: Follows best practices by instructing users to store API keys in a local file (~/.hypergen/credentials.env) with restricted filesystem permissions (chmod 600) rather than hardcoding them or using insecure temporary storage.
Audit Metadata