skills/alfredang/skills/project-init/Gen Agent Trust Hub

project-init

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill configures the agent to automatically load context from an external file (MEMORY.md) located in the user's home directory (~/.claude/projects/). This automated ingestion of data into the agent's prompt context creates a vector for indirect prompt injection attacks.
  • Ingestion points: The instructions generated in CLAUDE.md mandate that the agent 'Always read <MEMORY_PATH>/MEMORY.md at the start of each conversation for context from previous sessions.'
  • Boundary markers: There are no specific delimiters or instruction-ignore warnings defined to protect the agent from interpreting data within the memory file as commands.
  • Capability inventory: The skill facilitates file system operations including creating and modifying CLAUDE.md, .env, and directory structures in the user's home path.
  • Sanitization: The skill does not implement any validation or sanitization mechanisms for the content stored in the memory directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 07:29 AM
Security Audit — agent-trust-hub — project-init