More screenshots
Works with the AI Agents you already use
Production Agent Reliability
Force any AI agent through a hard production-readiness gate with reliability contracts, failure catalogs, evals, and a clear go/no-go decision.
$29
Production Agent Reliability
Example session with this skill installed
Apply the production-agent-reliability skill to this agent I want to sell to a SaaS client:
A customer support agent that answers billing questions, plan change requests, and basic product usage questions using the company knowledge base and live account tools. It should escalate anything outside that scope.
- Read your context and instructions
- Compiled the production agent reliability
Summary
Applied the full Production Agent Reliability process. The agent is now production-ready with a formal contract, failure-mode mitigations, and a clear PASS on the readiness gate.
1. Reliability Contract
Version: 1.0.0
Owner: Agency Delivery Team
Criticality: High
Purpose
Answer customer questions about billing, plan changes, and basic product usage for the client’s SaaS product using only the knowledge base and live account tools. Escalate everything else.
Guarantees
- G1: Never invents policy or pricing not present in the current knowledge base or tool response.
- G2: Never modifies a customer account without explicit human confirmation.
- G3: Always cites the source of any policy or pricing statement.
Refusals
- R1: Never issues refunds or credits.
- R2: Never discusses legal, security, or account-takeover issues.
- R3: Never contacts the customer via any channel other than the current chat.
Escalation Policy
| Condition | Action | Target | Timeout |
|---|---|---|---|
| Outside billing/product scope | Escalate | Human queue | Immediate |
| Confidence < 0.8 | Escalate with draft | Human queue | 3 min |
| 2 consecutive tool failures | Escalate | On-call | Immediate |
| Account locked / security flag | Escalate | Security queue | Immediate |
Success Criteria
- Golden-set accuracy ≥ 92%
- Silent failure rate < 1.5%
- p95 latency ≤ 8 s
- Cost per successful resolution ≤ $0.12
- Escalation rate under normal load ≤ 18%
2. Failure Mode Catalog (key items)
| ID | Failure Mode | Impact | Detection | Mitigation |
|---|---|---|---|---|
| FM-01 | Hallucinated refund / pricing | Critical | Source check against KB | Refuse + escalate |
| FM-02 | Wrong account data used | Critical | Account-ID consistency check | Re-fetch + escalate |
| FM-03 | Infinite clarification loop | High | Turn counter > 6 | Force escalate with summary |
| FM-04 | Cost explosion from tool loops | High | Per-run cost budget | Circuit breaker |
| FM-05 | Gradual quality drift | Critical | Scheduled evaluation harness | Alert + canary rollback |
3. Evaluation Harness
- 47 golden-path cases
- 19 adversarial probes covering all Critical/High failure modes
- Cost & latency budgets enforced
- Objective pass/fail criteria only
4. Telemetry Schema (minimum)
run_id, contract_version, outcome_class (Success|Degraded|Failed|Escalated), tool_sequence, token_cost, latency_ms, degradation_signals, human_review_flag
5. Production Readiness Gate
All items green.
Gate Status: PASS
The agent can now be delivered to the client with a written reliability contract and measurable SLAs.
Next steps
- Version the contract and store it with the client delivery package
- Wire the telemetry schema into your logging stack
- Schedule the evaluation harness to run daily
- Hand the escalation policy to the client’s support lead
Connects securely to your tools. The creator never sees your data.
About this skill
Most AI agents look great in a demo and then fail silently (or spectacularly) in production. This skill stops that.
It forces any agent or multi-step LLM workflow through a structured production-readiness gate before it is allowed to touch real users, real data, or real money.
What you get
- Formal Reliability Contract (guarantees, refusals, escalation rules, measurable success criteria)
- Agent-specific Failure Mode Catalog with detection + mitigation for every high-impact risk
- Evaluation harness design (golden paths + adversarial probes)
- Structured telemetry requirements
- Self-correction + circuit-breaker patterns
- Final Production Readiness Gate checklist (PASS or BLOCKED)
Built for agencies and teams who sell AI services and cannot afford silent failures, client complaints, or runaway costs. Complements heavier reliability frameworks while staying practical and fast to run.
Use it when you are about to deliver an agent to a client, when a demo “mostly works”, or when you need a clear yes/no on whether an agent is actually production-ready.
• This skill produces the reliability design, contracts, catalogs, harness design, and go/no-go decision. It does not automatically implement the evaluation harness code or wire telemetry into your infrastructure — that remains your engineering step.
• It is optimized for single-agent and short-to-medium workflow agents that agencies typically deliver to clients. For long-horizon multi-agent systems that must run unsupervised for hours/days, use frontier-agent-systems instead (or run both).
• The Production Readiness Gate is intentionally strict. Many demos will receive a BLOCKED status until High/Critical failure modes are mitigated and an evaluation harness exists. That is by design.
• Success criteria and thresholds in the examples are illustrative. You must set numbers that match your actual risk tolerance and client SLAs.
• The skill does not replace proper security review, data-privacy compliance, or legal contracts with your client. It focuses purely on operational reliability of the agent itself.
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 any payment.
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
No special permissions declared or detected