The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation guide instructs users to download a shell script from a third-party GitHub repository and execute it with bash, allowing for arbitrary code execution on the host system.
Evidence in SKILL.md: curl -fsSL https://raw.githubusercontent.com/CortexReach/toolbox/main/memory-lancedb-pro-setup/setup-memory.sh -o setup-memory.sh && bash setup-memory.sh.
[EXTERNAL_DOWNLOADS]: The skill depends on several external resources, including a remote setup script from raw.githubusercontent.com and the installation of the memory-lancedb-pro package from the npm registry.
[COMMAND_EXECUTION]: The skill makes extensive use of CLI-based management tools through the openclaw command, including system configuration changes, plugin installations, and database maintenance tasks like schema upgrades.
[CREDENTIALS_UNSAFE]: Setup requires the configuration of sensitive API keys for external services (OpenAI, Jina, and SiliconFlow). While instructions recommend environment variables, the skill specifically guides the handling of these credentials in its configuration files.
[PROMPT_INJECTION]: The skill implements a long-term memory system which introduces a surface for indirect prompt injection.
Ingestion points: Conversational data is automatically ingested via the autoCapture feature and manual memory_store tools as noted in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructional guardrails to prevent the agent from obeying commands embedded within recalled memory segments.
Capability inventory: The agent is granted capabilities to write to the memory database, update existing records, and extract new patterns or skills from untrusted input.
Sanitization: No sanitization or validation mechanisms are described for ensuring the safety of the content stored and subsequently retrieved from the vector database.

memory-lancedb-pro-openclaw