- Home
- Skills
- Agents & Orchestration
- AgentKit Rails Workflow Risk Check
AgentKit Rails Workflow Risk Check
Review an AgentKit Rails capability from caller identity through policy, approval, external execution, and receipt.
Free
Works with the AI Agents you already use
AgentKit Rails Workflow Risk Check
Example session with this skill installed
Hypothetical AgentKit Rails integration; not an audit of the AgentKit repository or any real application. Review this design for our charge_invoice capability: A worker constructs Agentkit::Principal from user_id and tenant_id supplied by the model. The worker calls the payment provider directly after the model chooses an invoice. A Start click is our approval. Sidekiq retries a timed-out charge, and we do not look up the provider result before retrying.
- Read your context and instructions
- Compiled the agentkit rails workflow
Needs changes — design review of a hypothetical AgentKit integration. These are stated design choices; I have not inspected code or run the workflow.
| Priority / ID | AgentKit path and evidence level | Risk or unknown | Smallest fix and owner | Verification test |
|---|---|---|---|---|
| High / AK-01 | Principal creation; Stated: IDs come from the model | The model can impersonate another actor or tenant | Rails app: derive actor and tenant from authenticated job context | Ask the model for another tenant; verify the principal stays bound to the caller |
| High / AK-03 | Worker to provider; Stated: direct provider call | This path can bypass AgentKit policy and action lifecycle | Rails app: route the operation through a governed capability/action path | Deny the permission; verify no provider call occurs |
| High / AK-04 | Approval; Stated: Start click | The approver never saw the final invoice and charge | Rails app: bind approval to exact action arguments before execution | Change the invoice after approval; verify execution is refused |
| High / AK-05 | Retry; Stated: no provider lookup before retry | A timeout after acceptance can duplicate a charge | Rails app: persist a stable provider key and reconcile unknown outcomes | Simulate acceptance then timeout; verify one charge |
Next actions: bind identity outside the model; use the governed action path; approve exact charge details; reconcile uncertain provider outcomes. The actual AgentKit configuration and provider behavior remain
Unknown until inspected.
Connects securely to your tools. The creator never sees your data.
About this skill
Review a capability or agent workflow built with AgentKit Rails. Eight checks trace how Principal and Context are derived, whether each entry point reaches Policy and Actions, how HITL approval binds to an action, and how an external result is reconciled after a retry. They also cover run_loop authorization and budgets, data handling, and receipts. Findings separate verified code, stated design choices, and unknowns, with a small fix assigned to AgentKit configuration or the host Rails app.
This is an Agensi review skill for developers using AgentKit, not an Agentkit::Skill installed inside the kernel. It does not execute workflows or certify a production deployment. For versions other than v1.12.0, inspect that version's source before making API-specific claims.
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.
Reviews only the workflow details or code you supply. It does not execute tests, inspect production, verify security controls, or certify compliance. Missing facts are marked Unknown.
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
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 13 days ago
- Free with an Agensi account
Permissions
No special permissions declared or detected