forge-observability

Warn

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command: node ../../../.fullstack-forge/runtime/cli/src/composition-entry.js observability compose --root <repository-root> --json. This executes a local script via a relative path traversal that reaches outside the skill's primary directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to ingest and strictly follow instructions from an external file (../../../.fullstack-forge/skills/forge-observability/SKILL.md) and a configuration file (.forge/composition.json). This creates a vulnerability surface where the agent's operational logic is determined by external files.
  • Ingestion points: The agent is instructed to read the canonical playbook and composition JSON file from the local filesystem based on relative paths.
  • Boundary markers: There are no explicit delimiters or safety instructions provided to the agent to treat the external content as untrusted data.
  • Capability inventory: The skill possesses command execution capabilities (using node and the forge CLI tools mentioned in the instructions).
  • Sanitization: No evidence of validation or sanitization of the externally loaded instructions is present.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 6, 2026, 06:48 PM
Security Audit — agent-trust-hub — forge-observability