scaffold-frenchfry-render

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill executes a local shell script (scripts/scaffold-frenchfry-tool-calling.sh) to automate project scaffolding. It also performs package installation via npm install for dependencies like @frenchfryai/runtime, hono, and react, which is a standard procedure for project initialization.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided directory names which are passed as arguments to shell scripts.
  • Ingestion points: Target directory name provided by the user in the scaffold command.
  • Boundary markers: None explicitly defined for the directory name argument.
  • Capability inventory: Shell script execution (bash), file writes (cat >), and package management (npm install) across the scaffolding process.
  • Sanitization: The underlying bash script uses standard safety flags (set -euo pipefail) and verifies directory existence before proceeding.
  • [SAFE]: The skill implements secure secret management by generating Render blueprints that use sync: false for the OPENAI_API_KEY environment variable, preventing the storage of sensitive credentials in version control systems.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 03:47 PM
Security Audit — agent-trust-hub — scaffold-frenchfry-render