smart-loader

Warn

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands to inspect the local filesystem (ls ~/.claude-context-optimizer/templates/) and runs a local Node.js script (tracker.js).
  • [COMMAND_EXECUTION]: The skill constructs a shell command using dynamic input: node ${CLAUDE_PLUGIN_ROOT}/src/tracker.js suggest "$(pwd)". Passing the current working directory via shell interpolation ($(pwd)) is a dynamic command generation pattern that could be exploited if the directory name contains shell metacharacters.
  • [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection by ingesting untrusted data (user-provided task descriptions) to influence which files are suggested for loading.
  • Ingestion points: User's task description (processed in SKILL.md).
  • Boundary markers: None explicitly defined for isolating the task description input.
  • Capability inventory: Shell command execution (ls, node).
  • Sanitization: The skill includes a manual safety check ("Do NOT automatically read files without user confirmation"), which mitigates the risk but does not eliminate the injection surface.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 23, 2026, 11:23 AM