byagent
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and manages the risk of malicious instructions being introduced via external feedback mechanisms.\n
- Ingestion points: Data from external users is ingested through the
byagent commentscommand and thebrief_getMCP tool, as described inSKILL.md.\n - Boundary markers: The skill contains clear, authoritative instructions to treat all comment text as untrusted data and specifically warns against treating comments as executable instructions.\n
- Capability inventory: The agent has access to terminal execution for the
byagentCLI, which allows for publishing, updating, and deleting hosted content.\n - Sanitization: The instructions explicitly restrict the agent to only applying content edits and strictly forbid the execution of shell commands, opening of links, or modification of credentials based on comment content.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes several external dependencies and resources during setup and execution.\n
- Fetches the
byagentutility from the public npm registry.\n - Instructs the user to clone and build the MCP server from the vendor's GitHub repository at
github.com/anup-a/artifacts.\n - Incorporates web assets for HTML artifacts from well-known providers, including Cloudflare's cdnjs, JSDelivr, and Google Fonts.\n- [COMMAND_EXECUTION]: The skill relies on terminal access to execute the
byagentCLI for authentication, publishing documents, and managing artifact lifecycles.
Audit Metadata