deploying-boxes

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill documents and utilizes command execution for administrative tasks.
  • The amp boxes exec tool is used to run production commands like database migrations or data imports, as seen in SKILL.md.
  • The scripts/configure-sign-in.sh script executes local shell commands to automate the registration of OAuth clients.
  • [CREDENTIALS_UNSAFE]: Hardcoded credential placeholders are present in development and testing files.
  • A test secret is used in template/tests/app.test.mjs for mocking authentication during the build process.
  • Instructions in SKILL.md and reference/sign-in-with-amp.md emphasize using amp secrets set with data files to ensure production secrets are never committed to the repository or visible in command history.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides an application template with an ingestion surface for user-generated content.
  • Ingestion points: User-submitted notes are processed in template/src/routes/+page.server.ts.
  • Boundary markers: The application is built using SvelteKit, which provides standard request handling and context boundaries.
  • Capability inventory: Agents utilizing this skill have the ability to execute remote commands via amp boxes exec and deploy code via amp boxes deploy.
  • Sanitization: Database interactions in template/src/lib/server/db.ts use node:sqlite with prepared statements to prevent injection attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 07:38 AM