hellowind777/helloagents

What it solves

HelloAGENTS addresses common failures in AI coding CLIs, such as agents stopping too early (providing suggestions instead of implementation), shifting responsibility to the user for hard tasks, losing project context across sessions, and reporting completion without actual verification. It adds a structured workflow layer to ensure AI agents act as capable executors that follow through to build, verify, and close out tasks.

How it works

It integrates as a layer on top of existing AI coding tools like Claude Code, Gemini CLI, Grok Build, Cursor, and Codex CLI. It uses a system of "skills" (specialized rule sets) and chat commands (prefixed with ~) to guide the agent's behavior.

Key mechanisms include:

• Workflow Skills: 14 built-in skills (e.g., hello-ui, hello-security, qa-review) that activate based on the task type to enforce quality standards.
• Project Knowledge Base: Maintains a .helloagents/ directory on disk to store project overviews, coding guidelines, and module notes, preventing the agent from having to rediscover facts.
• Structured Planning: Uses "plan packages" (requirements, tasks, and contracts) and PRD files to manage complex work instead of relying on chat history.
• State Recovery: Tracks session state in STATE.md and runtime.json to allow resumable execution across different workspaces or sessions.
• Verification Gates: Requires delivery evidence (like visual validation or test results) before a task is marked as complete.

Who it’s for

Developers using AI coding CLIs who want more reliability, stricter quality control, and better long-term project memory from their AI agents.

Highlights

• 14 Specialized Skills: Task-specific rules for UI, API, security, testing, architecture, and more.
• Chat-Driven Commands: Direct control via commands like ~plan, ~build, ~qa, and ~auto.
• Persistent Knowledge: On-disk storage for project context and guidelines to reduce context window waste.
• Delivery Evidence: Moves beyond "tests passed" to require structured evidence for final delivery.
• Broad Integration: Supports multiple major AI CLIs including Claude Code, Gemini, Grok, Cursor, and Codex.