forge-authorization

Fail

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a Node.js script located at ../../../.fullstack-forge/runtime/cli/src/composition-entry.js. This uses relative path traversal to execute code from a location outside the skill's root directory.
  • [DYNAMIC_EXECUTION]: The skill requires the agent to read and strictly follow instructions from a canonical playbook located at ../../../.fullstack-forge/skills/forge-authorization/SKILL.md. This bypasses skill isolation by loading instructions from the host file system.
  • [DYNAMIC_EXECUTION]: The skill instructs the agent to load ordered paths from a dynamically generated file .forge/composition.json at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where it ingests instructions from external files (SKILL.md via traversal and .forge/composition.json) and has command execution capabilities (node). It lacks sanitization, validation, or boundary markers to prevent malicious instructions in those external files from controlling the agent's behavior.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 6, 2026, 06:48 PM
Security Audit — agent-trust-hub — forge-authorization