forge-auth

Warn

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js script at a relative path: node ../../../.fullstack-forge/runtime/cli/src/composition-entry.js. This script is used to perform a "runtime composition" which determines how the skill behaves.
  • [PROMPT_INJECTION]: The skill identifies itself as an "adapter" that loads its primary instructions (the "canonical playbook") from an external file at ../../../.fullstack-forge/skills/forge-auth/SKILL.md. It also loads specialist guidance from paths listed in .forge/composition.json. This architecture creates a surface for indirect prompt injection if an attacker can influence the contents of these files within the repository.
  • Ingestion points: ../../../.fullstack-forge/skills/forge-auth/SKILL.md and .forge/composition.json relative to the repository root.
  • Boundary markers: Absent. The agent is told to "follow it exactly."
  • Capability inventory: The skill possesses the ability to execute Node.js scripts and read files throughout the repository structure via relative path traversal.
  • Sanitization: No sanitization or validation of the externally loaded instructions is mentioned in the adapter code.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 6, 2026, 06:48 PM
Security Audit — agent-trust-hub — forge-auth