Overview
Coding Pal is a centralized, version-controlled catalog of persistent context primitives designed to elevate AI coding assistants (GitHub Copilot, Claude Code, and Cursor) across engineering teams.
Maintained by ALTEN Group, Coding Pal solves the common failure modes of AI code assistants: hallucinated conventions, bloated token context, prompt drift, inconsistent architectural patterns, and unbounded agent behavior.
Key Capabilities
- Persistent Context Standardization — Defines engineering rules, conventions, and architectural boundaries in machine-readable markdown files that persist across sessions.
- Universal Multi-Harness Deployment — Authors guidance once in standard Markdown/YAML, and distributes automatically to GitHub Copilot, Claude Code, and Cursor using APM (Agent Package Manager).
- Sharp Agent Philosophy — Prevents overengineering, token waste, and hallucination through tightly scoped agents with strict boundaries and falsifiable completion criteria.
- The Golden Split — Separates task scope, universal coding rules, and machine-validated contracts into discrete, non-overlapping primitives.
- Scaffolding and Validation Skills — Bundles executable scripts, verification fixtures, and reusable templates alongside normative specifications.
The Problem Solved
Without a versioned persistent context catalog:
- Hallucinated Standards: Each developer re-explains company conventions (naming, layering, error codes, database migration practices) in every chat prompt.
- Context Window Waste: Dumping generic, repetitive guidelines into global system prompts rapidly depletes the model's context window.
- Prompt Drift: Teams diverge in coding standards because instructions live in local scratchpads or individual chat histories.
- Unbounded Agent Modification: AI agents frequently introduce unrequested dependencies, modify production code when only tests were requested, or declare success prematurely without running tests.
Coding Pal eliminates these issues by codifying standards into 4 well-defined primitives: Instructions, Agents, Skills, and Prompts.
Next Steps
- Understand the Persistent Context Architecture.
- Learn why Deterministic Controls are required to scale agent autonomy, and explore the Controls & Safeguards Domain.
- Explore Capabilities by Domain (Think & Plan, Testing, Auditing, Docs, Stacks, Controls).
- Browse Primitives by Type (Prompts, Agents, Instructions, Skills).
- Examine the exact Prompt Markdown Schema and Agent Markdown Schema.