Achilles Agents Library

Select a guide to explore Achilles Agents Library from architecture to hands-on tutorials.

Documentation Map

Use the links below to explore the major skill guides. Each page provides conceptual background, public API documentation, and practical guidance.

Skill System Overview

Learn how the new skill model decouples orchestration, execution, and evaluation while remaining LLM-friendly.

Backlog Manager

Manage backlog files, task lifecycles, approvals, and history moves with the BacklogManager API.

Agent Runtime

Understand how LLMAgent and RecursiveSkilledAgent collaborate to supervise multi-step plans.

Agentic Sessions

Deep dive into LoopAgentSession and SOPAgenticSession for stateful, multi-turn agent workflows.

Orchestration Skills

See how orchestration skills build loop/SOP agentic sessions, apply intent taxonomies, and monitor downstream execution.

LightSOPLang

Detailed syntax, semantics, and command mapping rules for the scripting language that powers agent plans.

LightSOPLang Monitoring

Understand default execution limits, monitoring hooks, and how to build custom observability for scripts.

DBTable Skills

See how table descriptors become CRUD workflows that plan operations, collect fields, and persist records.

Hands-on Tutorials

Step-by-step guides showing how to define skills, configure orchestration sessions, and integrate MCP tools.

MCP Skills

Discover how MCP-hosted toolchains are planned, scripted, and executed through the Agent runtime.

Code Skills

Generate executable code from detailed specifications and execute it safely.

Mirror Code Generator

Understand regeneration rules, test generation, and repair workflows for spec-driven code.

Claude Skills

Review static descriptors that surface skill summaries and bodies without running LLM actions.

Development Checklist

  1. Follow ES module semantics (import/export only).
  2. Use 4-space indentation, single quotes, trailing commas, terminated statements.
  3. Document public methods with JSDoc. Prefer descriptive error messages.
  4. Pair every change with tests. Run npm test before committing.
  5. Avoid destructive Git commands; preserve the user worktree.
  6. Read SKILLEDAGENT-ARCHITECTURE.md to align with existing patterns.