skills.rest
ruvnet
ruvnetCOMMUNITY
Explore all skills by this author →
Website𝕏@ruvGitHub

autopilot-predict

Predicts the optimal next action using learned patterns and current task state.

70.1k|8.4k|Updated Jun 2, 2025
Search Tags:#AgentDB#pattern learning#agent loop#MCP Tools#Workflow Automation#Autopilot#task prediction
One-click install
npx skills add https://github.com/ruvnet/claude-flow --skill autopilot-predict
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: autopilot-predict
Source: https://github.com/ruvnet/claude-flow/tree/main/plugins/ruflo-autopilot/skills/autopilot-predict
Command: npx skills add https://github.com/ruvnet/claude-flow --skill autopilot-predict

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Deciding what to work on next in an autonomous agent loop often requires manual prioritization and guesswork. This Skill uses learned patterns from past completions and current task state to recommend the next action, reducing decision overhead in long-running workflows.

Core Features & Use Cases

  • Confidence-Based Prediction: Calls the autopilot predict tool and executes the recommendation directly when confidence exceeds 0.7.
  • Task Fallback Logic: Falls back to the highest-priority incomplete task from the progress breakdown when prediction confidence is low.
  • Continuous Learning: Records completed work through the learn tool so patterns persist in AgentDB for cross-session recall.
  • Use Case: Inside a /loop iteration, the Skill guides each cycle—executing high-confidence predictions immediately, falling back to task priority order on low confidence, and exiting the loop when no tasks remain.

Quick Start

Ask the agent to predict what to work on next using the autopilot and execute the recommended action.

Frequently Asked Questions about autopilot-predict

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I predict the next task in an autonomous agent loop?▼

Call the autopilot predict tool to get a recommended next action with a confidence score. If confidence exceeds 0.7, execute it directly; otherwise check the progress tool and pick the highest-priority incomplete task.

How does the autopilot learn from completed tasks?▼

After completing work, call the autopilot learn tool to discover success patterns from finished tasks. Patterns are stored in AgentDB, enabling cross-session recall and history search by keyword.

What happens when prediction confidence is low?▼

When confidence falls below 0.7, the Skill checks the autopilot progress tool for a task breakdown and selects the highest-priority incomplete task instead of executing the prediction.

Can I search past task completions with the autopilot?▼

Yes, use the autopilot history tool with a query keyword to search past completions. This supports the learning pipeline by surfacing relevant prior work stored in AgentDB.

When does the autopilot loop stop running?▼

Inside a /loop, the Skill exits when no tasks remain in the progress breakdown. At that point autopilot is disabled and the loop terminates cleanly.

Related Skills

Explore More Software Engineering Skills→
ionclaw-orgionclaw-org
80

mcp-client

Connect to external MCP servers via Streamable HTTP transport to call tools and access resources.

Official
Intermediate
motionharvestmotionharvest
2

cas-operational-loop

Enforce an eight-phase operational loop with prediction, evaluation, and JSONL-based learning.

Community
Advanced
mindroom-aimindroom-ai
232

agent-cli-tmux-supervision

Supervise parallel agent-cli Codex worktrees with tmux sessions.

Official
Advanced
adaptive-enforcement-labadaptive-enforcement-lab
3

work-avoidance

Detect unnecessary work and skip execution in CI/CD pipelines.

Official
Intermediate