install-memory-powermem-full
Fail
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The installation guide instructs users to download and execute a shell script directly from the vendor's GitHub repository using
curl -fsSL ... | bash. While this originates from the official repository of the skill author, piping remote scripts to a shell bypasses local package management and manual code review. - [EXTERNAL_DOWNLOADS]: Fetches installation scripts and configuration templates from the vendor's public repository on GitHub.
- [COMMAND_EXECUTION]: The skill requires the agent to execute local CLI commands such as
pmem,openclaw, andpython3to manage the memory plugin and verify system requirements. - [PROMPT_INJECTION]: The skill implements a long-term memory system where the agent can search and retrieve past interactions. This introduces an indirect prompt injection surface, as the retrieved content is injected into the model's context via a
<relevant-memories>block, potentially allowing malicious data stored in memory to influence the agent's behavior. - Ingestion points: Data retrieved from the PowerMem database via the
memory_recalltool as described inSKILL.md. - Boundary markers: The skill mentions that memories are injected using a
<relevant-memories>block, which serves as a delimiter but does not prevent instruction obedience within the block. - Capability inventory: The skill allows for storing, searching, and deleting memory entries, and executes the
pmemCLI tool as a subprocess. - Sanitization: No specific sanitization or validation of the retrieved memory content is mentioned before it is injected into the context.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/ob-labs/memory-powermem/main/install.sh - DO NOT USE without thorough review
Audit Metadata