Imported from Charlesganu2004/Agentic-Systems-Blueprints (docs/AGENTS.md). Install upstream with npx skills add Charlesganu2004/Agentic-Systems-Blueprints --skill docs. Copyright stays with the author.

Building Agents

Name: docs
Author: Charlesganu2004

An agent is useful when the system must choose steps, call tools, inspect results, and decide what to do next. If the task is just "answer from documents," start with RAG before adding agents.

Agent Design Template

Every agent should have:

Name
Role
Outcome
Inputs
Tools
Boundaries
Escalation rules
Review requirements
Output format
Success metric

Common Agent Types

Agent	Job
Orchestrator	Routes work, keeps the user informed, resolves conflicts.
Researcher	Finds and summarizes source-backed information.
Architect	Designs system boundaries and technical choices.
Builder	Implements scoped code or configuration.
Tester	Verifies behavior and writes test evidence.
Auditor	Checks safety, legal, privacy, correctness, and process evidence.
Prompt Efficiency	Reduces context waste before expensive model calls.

Good Boundaries

Agents should not own secrets.
Agents should not skip review on legal, financial, privacy, or production changes.
Agents should not invent sources.
Agents should prefer small, verifiable tasks.
Agents should write durable evidence when work affects a repo.

Implementation Options

Markdown charters in the repo.
JSON/YAML agent registries.
LangGraph-style workflow graphs.
Simple orchestrator code that calls named tools.
GitHub issue-driven work queues.
Human approval gates for expensive or high-risk steps.

{ "ocm": "1", "id": "charlesganu2004-agentic-systems-blueprints-docs", "kind": "skill", "name": "docs", "description": "An agent is useful when the system must choose steps, call tools, inspect results, and decide what to do next. If the task is just \"answer from documents,\" start with RAG before adding agents.", "publisher": "Charlesganu2004", "version": "1.0.0", "capabilities": { "domains": [ "general" ], "tags": [ "skill-md", "github-agents" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "An agent is useful when the system must choose steps, call tools, inspect results, and decide what to do next. If the task is just \"answer from documents,\" start with RAG before adding agents." ], "primary": false, "metadata": { "source": { "provider": "github-agents", "repository": "https://github.com/Charlesganu2004/Agentic-Systems-Blueprints", "path": "docs/AGENTS.md", "ref": "6916f958734d7c1d81daf3635711d2fce6cf70ee", "url": "https://github.com/Charlesganu2004/Agentic-Systems-Blueprints/blob/6916f958734d7c1d81daf3635711d2fce6cf70ee/docs/AGENTS.md", "key": "Charlesganu2004/Agentic-Systems-Blueprints/docs/AGENTS.md" } }, "instructions": "# Building Agents\n\nAn agent is useful when the system must choose steps, call tools, inspect results, and decide what to do next. If the task is just \"answer from documents,\" start with RAG before adding agents.\n\n## Agent Design Template\n\nEvery agent should have:\n\n- Name\n- Role\n- Outcome\n- Inputs\n- Tools\n- Boundaries\n- Escalation rules\n- Review requirements\n- Output format\n- Success metric\n\n## Common Agent Types\n\n| Agent | Job |\n|-------|-----|\n| Orchestrator | Routes work, keeps the user informed, resolves conflicts. |\n| Researcher | Finds and summarizes source-backed information. |\n| Archite", "cost": { "context_tokens": 370 } }

docs

About

Building Agents

Agent Design Template

Common Agent Types

Good Boundaries

Implementation Options

Use it

Manifest

Reviews