Repository context
AGENTS.md carries durable project commands and conventions across a growing set of coding agents.
Learning center · Updated July 2026
Build agents that hold up in production
Start with the workflow, document the boundaries, connect the right tools, and verify the result. These guides favor reusable operating knowledge over one-off prompt tricks.
Current signals
The agent stack is separating into clear layers
The most useful 2026 shift is not another model release. It is the emergence of portable formats for instructions, workflows, tools, agent interoperability, and commerce.
8 of 8 monitored primary sources verified · Latest source activity Jul 24, 2026
Reusable procedures
Agent Skills package on-demand workflows in SKILL.md files. The official repository monitor is at revision 38a2ff82958a.
Tools and data
MCP 2025-11-25 is the monitored stable release; 2026-07-28-RC remains a preview. Authentication and independent security review remain essential.
Agent interoperability
A2A v1.0.1 is the current monitored stable release for Agent Cards, tasks, messages, artifacts, and protocol bindings.
Agent commerce
UCP covers the shopping journey and can pair with AP2 when verifiable payment authorization is required.
Choose a learning path
Each path starts with examples you can inspect, adapt, and validate against your own stack.
Learn each model on its own terms
Compare the context, reasoning, tool, and prompting patterns that make Claude, GPT, Gemini, Grok, and DeepSeek more reliable.
Explore model learning curvesPackage reusable agent skills
Turn a repeatable procedure into a portable SKILL.md playbook with optional scripts, references, and assets.
Learn the Agent Skills formatWrite durable agent instructions
Turn repository knowledge, commands, constraints, and review standards into a focused AGENTS.md file.
Browse AGENTS.md examplesConnect agents to real tools
Compare MCP servers, understand their prerequisites, and review the data or actions each integration exposes.
Explore MCP serversScope Cursor project rules
Learn when a rule should always apply, attach to a file pattern, or stay available for manual use.
Browse Cursor rulesDesign an agent-to-agent workflow
Document triggers, actions, ownership, approvals, and failure handling before connecting two agents.
Explore A2A workflowsA reliable workflow
Context before automation
An agent is easier to trust when its inputs, permissions, approval points, and success criteria are explicit.
- 01
Define the job
State the outcome, owner, inputs, and measurable completion condition.
- 02
Constrain the agent
Document allowed tools, sensitive data, approval gates, and prohibited actions.
- 03
Test the full path
Verify the happy path, expected failures, retries, and the final user-visible result.
Verify against primary sources
Community examples are useful starting points. Product behavior and protocol details should be checked against their maintainers.
Agent Skills specification
The open SKILL.md format for portable, on-demand workflows.
AGENTS.md
Cross-tool repository instructions and nested-file behavior.
Cursor Rules
Current rule locations, types, scoping, and examples.
Claude Code memory
How CLAUDE.md files load and how to write effective instructions.
Model Context Protocol
The official MCP introduction, architecture, and SDK guides.
A2A protocol 1.0
The current specification for interoperable agent discovery and tasks.
Latest from the field
2026-07-26
Agent Architecture Decision Memo: Subagents vs Automations + MCP vs A2A
A source-backed decision memo for Subagents vs Automations and MCP vs A2A.
2026-07-26
Agent Stack Brief: Automations, Evals and verification, Permissions and safety, Agent commerce and payments
A primary-source field brief on Automations, Evals and verification, Permissions and safety, Agent commerce and payments, with decision boundaries and production checks.
2026-07-26
Agent Systems Knowledge Brief: Tasks, Memory and context, Context engineering, Tools and MCP, Multi-agent systems, Agent interoperability
A source-backed weekly rotation through Tasks, Memory and context, Context engineering, Tools and MCP, Multi-agent systems, Agent interoperability, with decision rules, implementation questions, artifacts, and lifecycle checks.