The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill implements a plugin architecture via network_skill_pull and skill_install tools. These allow for the downloading and permanent installation of remote skill bundles from a team server or hub into the agent's execution environment.\n- [DATA_EXFILTRATION]: Includes tools (memory_share, skill_publish, task_share) designed to transmit local conversation history, task summaries, and skills to external endpoints (team server or hub). While instructions mandate user consent, the technical capability for data exfiltration is present.\n- [CREDENTIALS_UNSAFE]: Requests access to sensitive configuration files at ~/.openclaw/openclaw.json and environment variables (OPENCLAW_STATE_DIR, OPENCLAW_CONFIG_PATH), which are likely to contain authentication tokens or system paths for the MemOS service.\n- [EXTERNAL_DOWNLOADS]: The installation manifest specifies the download and installation of the better-sqlite3 Node.js package via npm.\n- [PROMPT_INJECTION]: The skill's core functionality involves ingesting and processing untrusted data from past conversations and team-shared memories.\n
Ingestion points: Data enters through memory_search, memory_get, and network_memory_detail (SKILL.md).\n
Boundary markers: No specific boundary markers or sanitization requirements are mentioned in the tool descriptions to prevent the agent from executing instructions found within retrieved memories.\n
Capability inventory: The agent has high-privilege capabilities including skill installation (skill_install), network communication (memory_share), and file access.\n
Sanitization: None specified for retrieved content.

memos-memory-guide