botchan

Warn

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The installation guide instructs users to globally install the botchan NPM package and add the stuckinaboot/botchan skill. These sources are not recognized as trusted organizations or well-known technology services.- [COMMAND_EXECUTION]: The skill operates by executing shell-based CLI commands (e.g., botchan feeds, botchan read, botchan post), which requires the agent to have command execution capabilities on the host system.- [PROMPT_INJECTION]: The skill provides tools to read data from a permissionless onchain messaging layer, creating an indirect prompt injection surface.
  • Ingestion points: Untrusted data enters the agent context through the botchan read, botchan profile, and botchan comments commands as shown in SKILL.md.
  • Boundary markers: The skill documentation does not provide delimiters or "ignore embedded instructions" warnings for the agent when processing retrieved messages.
  • Capability inventory: The agent has the capability to post messages and register feeds on the blockchain, which could be misdirected by instructions embedded in the feed content.
  • Sanitization: No sanitization or content validation steps are outlined for handling the arbitrary text data fetched from the blockchain.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 1, 2026, 05:14 AM