memory
Fail
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The
memory.shscript executes shell commands and invokespython3to manage data storage. - [REMOTE_CODE_EXECUTION]: The script
memory.shcontains a critical code injection vulnerability. It interpolates shell variables directly into Python heredocs (e.g.,value = """$VALUE"""). Because these variables contain unsanitized user input, an attacker can use triple quotes to terminate the Python string and execute arbitrary Python code, which can then be used to execute shell commands. - [DATA_EXFILTRATION]: Although the skill is designed for local storage in
~/.agent-memory/memory.json, the code injection flaw allows an attacker to read this file or any other sensitive file on the system and potentially transmit the contents over the network.
Recommendations
- AI detected serious security threats
Audit Metadata