remember

Warn

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (grep) using user-provided keywords to check for duplicates in MEMORY.md. If the input contains shell metacharacters (such as ;, &, |, or backticks), it could lead to arbitrary command execution on the host system.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input and persists it in a file that the agent is designed to read automatically at the start of every session.
  • Ingestion points: The <what to remember> user input is captured and written to MEMORY.md as described in Step 3 of SKILL.md.
  • Boundary markers: None. User input is appended directly to the markdown file without delimiters or instructions to the agent to ignore embedded commands.
  • Capability inventory: The skill uses shell commands (grep, sed) and file system writes to manage the memory file.
  • Sanitization: No sanitization, escaping, or validation is performed on the user-provided text before it is used in shell commands or written to the memory file.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 3, 2026, 02:33 PM
Security Audit — agent-trust-hub — remember