The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its reliance on external task history and automated predictions.
Ingestion points: The skill retrieves data via mcp__claude-flow__autopilot_predict and mcp__claude-flow__autopilot_history.
Boundary markers: No boundary markers or delimiters are defined to isolate untrusted data from the task history.
Capability inventory: The skill instructions grant the agent the capability to execute instructions derived from tool outputs.
Sanitization: There is no evidence of validation or sanitization of the patterns stored in AgentDB.
[COMMAND_EXECUTION]: The skill promotes high autonomy by instructing the agent to bypass human review for certain actions.
Evidence: The instruction to "execute the prediction directly" if confidence is greater than 0.7 allows for tool-driven command execution without explicit user confirmation.

autopilot-predict