Mantis Productivity Plugin: Anti-Stall Behavioral Correction System for AI-Assisted Development

A Claude Code plugin designed to prevent common AI assistant failure patterns during software development, specifically addressing the "70% completion stall" where AI assistants systematically derail projects approaching completion. ## Problem Addressed Empirical analysis of 100+ development conversations (October 2025 - January 2026) revealed consistent behavioral patterns where AI assistants: - Ask unnecessary clarifying questions at ~70% project completion - Propose alternative architectures instead of executing deployment - Generate analysis documents instead of working code - Create decision paralysis through excessive option presentation ## Solution Components **12 Slash Commands:** - /ship - Force immediate deployment - /execute - Code-only output mode - /no-questions - Autonomous decision mode - /finish - Complete final 30% without stalling - /70-check - Self-audit for stall patterns - /status - Project completion dashboard - /blob, /continue - Session persistence - /unblock - Blocker resolution - /deploy-netlify, /deploy-github - Direct deployment **3 Specialized Agents:** - executor - Zero-architecture code production - deployer - Deployment automation - completion-tracker - Progress monitoring **3 Behavioral Hooks:** - pre-response-stall-check - Intercept stall patterns - token-limit-blob - Auto-save session state - post-task-completion-tracker - Progress logging **2 Skills:** - anti-architecture-addiction - Prevent over-engineering - deployment-first - Prioritize shipping over perfection ## Innovation Claims 1. First systematic documentation of AI assistant "completion stall" patterns 2. Behavioral correction through slash command enforcement 3. Hook-based response interception for pattern prevention 4. Session state serialization for cross-conversation continuity This publication establishes prior art for AI productivity correction methodologies.