mycelium

Fail

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructions advise users to install components by piping a remote script directly to bash: curl -fsSL https://raw.githubusercontent.com/openprose/mycelium/main/install.sh | bash. This execution pattern is insecure as it runs unverified code from a remote source with the user's local privileges.\n- [COMMAND_EXECUTION]: The skill configuration includes the registration of several automated hooks (PostToolUse, SessionStart, Stop, SubagentStart) that execute shell and Python scripts automatically during the agent's operation, establishing a persistent execution environment for external tools.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its design of surfacing external git notes as context for the agent.\n
  • Ingestion points: Data is ingested from refs/notes/mycelium metadata via the mycelium.sh tool and injected into the agent's context as additionalContext.\n
  • Boundary markers: No delimiters or safety warnings are specified to separate the retrieved notes from the agent's instructions.\n
  • Capability inventory: The skill leverages automated hooks and shell scripts that can modify the file system and execute git commands.\n
  • Sanitization: There is no indication that the contents of the git notes are sanitized or validated before being provided to the agent.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/openprose/mycelium/main/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 29, 2026, 11:46 PM