Works with the AI tools you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsagent handoff contract generator
Audit and harden data contracts between agents in OpenAI Agents SDK, CrewAI, and LangGraph to prevent silent corruption.
$14
agent handoff contract generator
Example session with this skill installed
I have a LangGraph setup where nodes update the state using Command(update=...). Sometimes downstream nodes get missing fields and crash. Can you audit my state handoff?
- Read your context and instructions
- Compiled the agent handoff contract
- Generated the document
I have audited the LangGraph state handoff in your provided graph definition. The primary issue is the use of `TypedDict` for your `AgentState`, which provides static type hints but zero runtime validation. When an upstream node omits a field in `Command(update=...)`, the graph continues with a corrupted state, causing the downstream crashes you observed. ## Handoff Contract Audit | Location | Framework | Finding | Severity | | :--- | :--- | :--- | :--- | | `AgentState` | LangGraph | `TypedDict` used; no runtime enforcement | High | | `research_node` | LangGraph | Missing `schemaVersion` in update payload | Medium | | `writer_node` | LangGraph | Missing `status` field for outcome tracking | Medium | ## Hardened State Contract I have converted your `TypedDict` to a Pydantic `BaseModel` and added versioning. This ensures the graph raises a `ValidationError` immediately at the handoff point if a node returns malformed data. ```python from pydantic import BaseModel, Field from typing import List, Optional class AgentState(BaseModel): # Added versioning and status for contract safety schema_version: str = Field(default="1.0.0") status: str = Field(default="pending") # Your existing fields now with runtime enforcement query: str documents: List[str] = Field(default_factory=list) draft: Optional[str] = None # Update your node logic to return a validated dict # Example for research_node: # return Command(update=AgentState( # schema_version="1.0.0", # status="success", # query=state.query, # documents=found_docs # ).model_dump()) ``` **Next steps** 1. Replace the `TypedDict` definition of `AgentState` with the Pydantic `BaseModel` provided above. 2. Initialize your graph with the new `AgentState` model. 3. Update node return statements to use `.model_dump()` to ensure the payload is validated before the state update occurs.
agent-handoff-contract-generator.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
In multi-agent systems, data corruption happens when an upstream agent changes its output shape and a downstream agent silently receives the garbage. This skill audits handoff points in OpenAI Agents SDK, CrewAI, and LangGraph codebases to find these silent failure vectors. It identifies missing schemas and provides the framework-specific code to enforce runtime validation at the exact moment data crosses between agents. ## What it does - **Handoff auditing** scans code for `handoff()` calls, CrewAI `Task` definitions, and LangGraph `Command` updates missing explicit contracts. - **Contract generation** creates Pydantic models or typed schemas that include mandatory `status` and `schemaVersion` fields. - **Runtime enforcement** provides code that fails loud on malformed payloads, preventing silent state corruption. - **Framework-specific wiring** generates diffs for the buyer's actual call sites using the native mechanisms of their chosen SDK. ## How it works 1. **Framework identification** detects if the codebase uses OpenAI Agents SDK, CrewAI, or LangGraph. 2. **Handoff inventory** maps every point where data passes between agents, tasks, or nodes. 3. **Gap reporting** lists every handoff point missing runtime validation, versioning, or status fields. 4. **Validator implementation** generates the missing framework-native code to harden the data contract. ## Frameworks & tools This skill supports OpenAI Agents SDK, CrewAI, and LangGraph. It uses Pydantic for runtime validation and targets the native handoff mechanisms of each framework. ## Why this beats prompting it yourself Generic prompts often suggest framework-agnostic JSON schemas that require custom boilerplate to implement. This skill understands the specific internal validation hooks of each SDK and identifies the specific risk of using `TypedDict` in LangGraph for runtime safety. ## Use cases - Auditing a CrewAI project where unstructured text is passed between tasks without a schema. - Hardening OpenAI Agents SDK handoffs to ensure every transfer includes a status code. - Converting LangGraph `TypedDict` states into Pydantic models for strict runtime enforcement. - Root-causing why a downstream agent is failing due to malformed upstream data. ## Known limitations Does not audit control flow for deadlocks or infinite loops. Only supports OpenAI Agents SDK, CrewAI, and LangGraph. Requires access to the buyer's actual call-site code for accurate diffs.
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 1 day ago
- Passed all security checks, Safe to install