The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to run npx @claude-flow/cli@latest, which downloads and executes code from the public NPM registry at runtime. The use of the @latest tag pulls the most recent version of the tool without version pinning, posing a risk of executing unvetted or malicious updates.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute npx @claude-flow/cli@latest swarm watch --stream, a persistent process whose output is monitored and processed by the agent.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it processes live swarm event data which may contain untrusted content from the environment.\n
Ingestion points: The stdout line events produced by the swarm watch --stream command in SKILL.md.\n
Boundary markers: No delimiters or specific instructions are provided to help the agent distinguish between trusted commands and untrusted data within the stream.\n
Capability inventory: The agent has access to the Bash tool and multiple MCP tools (mcp__claude-flow__swarm_status, mcp__claude-flow__swarm_health), which could be targeted by instructions embedded in the monitored data.\n
Sanitization: There is no evidence of sanitization or schema validation for the NDJSON events before they are processed by the agent.

monitor-stream