Claude Code Guide
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Prompt Injection (LOW): The skill suggests instructing the agent to "forget previous instructions" as a debugging technique. This is a common pattern used in malicious prompt injections to bypass or reset system-level constraints and safety guidelines.
- Indirect Prompt Injection (LOW): The workflow promotes reading and following instructions from a project-root
CLAUDE.mdfile. - Ingestion points:
CLAUDE.mdfile in the root directory of a project. - Boundary markers: Absent; the provided template for
CLAUDE.mddoes not include delimiters or instructions for the agent to ignore embedded commands. - Capability inventory: The skill suggests specific commands for execution, such as
npm run dev,npm test, andnpm run buildbased on the file's contents. - Sanitization: Absent; the agent is expected to read the markdown file and directly adopt the commands and workflow described within.
Audit Metadata