- Home
- Skills
- Agents & Orchestration
- Agent Routing Brain — Choose the Best AI Route
Works with the AI Agents you already use
Agent Routing Brain — Choose the Best AI Route
Match each task to the right model, tool, data source, workflow, or human checkpoint.
$9.99
Agent Routing Brain — Choose the Best AI Route
Example session with this skill installed
Plan a route for this task: summarize five uploaded contracts, extract renewal dates, compare them with our policy checklist, and produce a review table. Available capabilities: file reading, a general-purpose LLM, a deterministic date parser, and a human legal reviewer. Documents may contain confidential information. Accuracy is more important than speed. Do not assume network access.
- Read your context and instructions
- Compiled the agent routing brain
Recommended Task Routing Plan
Summary
Use local file reading for the supplied contracts, a language model for clause extraction and plain-language summaries, a deterministic parser to normalize dates, and a human legal reviewer for policy interpretation and final approval. No external browsing or network access is assumed.
Task graph
| Step | Route | Input | Output / check |
|---|---|---|---|
| 1. Inventory | File reader | Five supplied files | Confirm filenames, readability, page count, and any missing files |
| 2. Extract clauses | LLM with focused prompts | Relevant contract text | Renewal, notice, term, and termination language with page/section evidence |
| 3. Normalize dates | Deterministic date parser | Quoted date expressions and reference date | ISO date, timezone/context if known, and parse status; flag ambiguity |
| 4. Compare policy | LLM against supplied checklist | Evidence-backed extracted terms and policy checklist | Match, mismatch, unclear, or missing; cite contract evidence |
| 5. Review | Human legal reviewer | Draft comparison table and source excerpts | Confirm interpretations, resolve ambiguity, approve final output |
| 6. Deliver | LLM formatting pass | Reviewer-approved facts | Concise summary and traceable table; no unreviewed legal conclusion |
Data and safety controls
Keep processing within the approved environment. Do not send contract content to a network service unless the user explicitly authorizes that specific destination. Minimize copied excerpts, avoid retaining unnecessary personal data, and label any missing or unreadable content.
Validation and fallback
Check every extracted date against its quoted source text. If a parser cannot normalize a date, preserve the original wording and mark it “needs review.” If the model finds no renewal clause, distinguish “not found in supplied text” from “no renewal term.” If files are unreadable or policy is missing, stop that comparison step and report the gap rather than guessing.
Decision rationale
Accuracy is the priority, so deterministic parsing handles date normalization and a human reviewer handles legal interpretation. The LLM assists with extraction and organization but does not replace legal review.
Next steps
Confirm the processing environment and approved file scope, supply the policy checklist, then produce a draft table with document name, renewal date, notice period, cited clause, policy status, confidence, and reviewer notes.
Connects securely to your tools. The creator never sees your data.
About this skill
Agent Routing Brain
Choose a suitable route for each part of a task: model, tool, data source, deterministic function, workflow, human checkpoint, or fallback. The goal is to meet the required quality with an appropriate balance of safety, speed, reliability, freshness, and cost.
What it does
- Breaks a request into steps and identifies capability requirements and dependencies.
- Compares available routes using user constraints, risk, freshness needs, and measured evidence.
- Recommends model, tool, connector, retrieval, and human-review assignments.
- Defines order, inputs, validation, fallback, escalation, and stop conditions.
- Flags unknown capabilities and information that must be verified before execution.
How to use
Describe the task, the tools and models that are actually available, required quality, latency or cost constraints, and any safety boundaries. The agent returns a transparent routing plan, explains tradeoffs, and identifies what must be checked at runtime. It does not invent integrations, provider availability, prices, permissions, or successful actions.
Deliverables
Task graph, capability checklist, route comparison, recommended sequence, guardrails, validation steps, fallback plan, and decision log. Suitable for single-model workflows, tool-using agents, RAG, APIs, MCP systems, browser tasks, and multi-agent orchestration.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Recommendations depend on the accuracy of the stated task constraints and available-capability inventory. Provider features, availability, pricing, latency, permissions, and performance can change and must be verified before execution. This skill does not execute routes.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Permissions
Uses only the capabilities and tools explicitly available or supplied. Does not connect to services, route or execute a task, or validate provider performance on its own. Treat data-handling recommendations as planning guidance and verify platform permissions before use.