Agents Best Practices skill
agents-best-practices
A provider-neutral skill for designing agent systems: loop architecture, tools, permissions, memory and observability
- 2,405 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill consists only of instructions and references
Install
- Works in
Manual install
npx skills add DenisSergeevitch/agents-best-practices -gThe -g flag installs the skill at user level for all projects.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill consists only of instructions and references
This is third-party code. Review the repository files before installing.
What it does
The skill helps design, scaffold an MVP for, audit, refactor and explain agent harnesses for any domain. It sets a base architecture: the model proposes an action, the harness validates the schema, decides on permissions, executes or pauses for approval, returns a structured observation and updates context. Coverage spans tool design, loop budgets, context compaction, memory, skills, external connectors, observability, evals and safety. The approach is provider-neutral and fits not only coding agents but research, support, operations, finance, legal and other workflows.
Who it is for. For engineers and product teams who design and review agent systems.
Good fit when
- You design an agent harness for a new domain
- You audit an existing agent for reliability, cost and safety
- You choose between direct tool loops, hosted tools and SDKs
Not a fit when
- You need plain single-turn writing, translation or Q&A
- The task is not about agent design
Example request
Design an MVP harness for an agent that reads the CRM and drafts emails, with send approvalLimitations
The skill gives architectural principles and blueprints, not finished harness code for a specific framework. The material is instructions and references only.
How to disable. Remove the agents-best-practices folder from your agent's skills directory.
FAQ
Only for coding agents?
No, the principles apply to research, finance, legal, operations and other agents.
Is the skill tied to one API?
No, it is provider-neutral and covers OpenAI, Anthropic and compatible APIs.
Discussion
No questions or reviews yet.
Related
SkillsSkills on designing agent systems: context management, multi-agent patterns, memory, tools, evaluation and harnesses
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A skill plus hooks that keep a coding agent on task and cut speculative defenses, unread checksums and scope creep
A parallel divergent ideation skill for agents: fans out under cognitive frames, prunes traps and deepens the best branches
A completion-discipline skill for large agent tasks: write the acceptance ledger with runnable checks first, then do the work, then reverify against evidence
A set of CLI tools and a skill that give a text-only agent sight: image analysis, long-screenshot OCR and UI restoration
README in short
The README presents a provider-neutral skill for designing, scaffolding an MVP for, auditing and refactoring agent harnesses. Installation is via npx skills with a global flag, a ready prompt for the agent, or manual cloning into the skills directory. It offers seven common cases, from generating a domain blueprint to auditing an existing agent with a fix order. The core idea: the model proposes actions and the harness validates, authorizes, executes and records them. MIT licensed.
README badge
Original SKILL.md
--- name: agents-best-practices description: "Use this skill when designing, generating an MVP blueprint for, auditing, refactoring, or explaining an agentic harness for any domain. Covers provider-neutral agent architecture for OpenAI, Anthropic, and OpenAI-compatible APIs: agent loops, tool design, record provenance, interactive presentation, user-memory lifecycles, environment-adaptive tools, speculative tool execution, late-bound capabilities, permissions, system prompts, planning, goals, context compaction, memory, skills, MCP/external connectors, public-board communications, self-refining recursive harnesses, programmable context, continual refinement, observability, evals, prompt caching, agent-legible environments, feedback loops, and safety." metadata: version: "1.7.0" scope: "provider-neutral-agent-harness" file_policy: "markdown-only" --- # Agents Best Practices Use this skill when the user asks how to build, improve, debug, or evaluate an agentic harness.