skills/tingyulu/myr2d2/pickup/Gen Agent Trust Hub

pickup

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill demonstrates a surface for indirect prompt injection by processing external data as instructions.
  • Ingestion points: The skill reads the full content of Markdown files located in .claude/handoffs/ into the active session context.
  • Boundary markers: None are specified; the agent is instructed to read the "full text" including context and completion definitions without delimiters to isolate untrusted content.
  • Capability inventory: The skill uses shell commands (grep) and file editing tools to manage the handoff lifecycle.
  • Sanitization: No validation or sanitization is performed on the content of the handoff files before they are interpreted as working instructions.
  • [COMMAND_EXECUTION]: The skill executes a local shell command (grep -l "^status: pending" .claude/handoffs/*.md) to identify tasks. This is restricted to a specific local directory and uses a static pattern, which is considered a standard operational practice for this use case.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 10:37 AM
Security Audit — agent-trust-hub — pickup