The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation provides a CLI alternative using 'npx @claude-flow/cli@latest', which downloads and executes the vendor's package from the npm registry at runtime.
[COMMAND_EXECUTION]: The skill configuration allows the 'Bash' tool, and the instructions include several shell command examples for initializing and searching the vector database.
[PROMPT_INJECTION]: The skill implements search operations like 'embeddings_search' and 'memory_search_unified' that process arbitrary query input. Because the agent has access to powerful tools like 'Bash', this creates an indirect prompt injection surface where untrusted content retrieved during a search (the corpus data) could potentially provide instructions that the agent attempts to follow. Evidence: 1. Ingestion points: 'embeddings_search' and 'memory_search_unified' tools in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Bash access and multiple MCP tool invocations. 4. Sanitization: None described.

vector-search