slb

Fail

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The installation instructions recommend downloading a shell script from a remote repository and piping it directly to bash. This allows for unverified code execution with user-level privileges.\n- [COMMAND_EXECUTION]: The skill is primarily a wrapper for executing arbitrary shell commands, which are classified into risk tiers. It exposes commands like slb run and slb emergency-execute to the agent.\n- [PERSISTENCE]: The skill provides tools to modify agent-specific configuration files (e.g., ~/.claude/hooks.json and .cursorrules) to integrate with the tool's gating mechanism, effectively persisting its presence in the agent's workflow.\n- [INDIRECT_PROMPT_INJECTION]: The system's 'two-person rule' relies on external input (reviews/approvals) that could be manipulated by an attacker who influences the agent's context or the review chain.\n
  • Ingestion points: Command strings passed to slb run in SKILL.md.\n
  • Boundary markers: None identified in the provided documentation for command inputs.\n
  • Capability inventory: Full shell command execution via the slb tool documented in COMMANDS.md.\n
  • Sanitization: Relies on risk patterns rather than strict sanitization of the command strings.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/Dicklesworthstone/slb/main/scripts/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Nov 4, 2026, 09:51 AM