Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    reliability failure semantics reviewer

    1

    Senior principal-level review of agent reliability, failure semantics, recovery design, and operational bounds. Outputs severity-ranked findings only.

    $45

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    reliability failure semantics reviewer

    reliability failure semantics reviewer

    Example session with this skill installed

    Review the reliability and failure semantics of the following agent system:

    The agent calls an LLM in a while True loop with no maximum retry count. Tool calls (web_search, db_query, file_read) have no timeouts. Session state is written with a plain json.dump after each step. On parse failure the agent returns None and continues. Recovery retries a failed step 5 times then retries 5 more times with no human escalation.

    • Read your context and instructions
    • Compiled the reliability failure semantics

    Critical

    • Unbounded retry on LLM API calls with no circuit breaker
      Evidence: while True loop with no maximum attempt count.
      Impact: Indefinite retry on provider outage consumes quota and blocks the worker.
      Required fix: Bounded retry (max 3), exponential backoff with jitter, circuit breaker after 5 consecutive failures.

    • State corruption on crash mid-workflow with no transactional writes
      Evidence: json.dump used without atomic write.
      Impact: Crash leaves partially written session state. Restart may proceed on corrupted data.
      Required fix: Atomic write (temp file + replace) + schema validation on load.

    High

    • Timeout configured only on main LLM call, not on tool sub-calls
      Evidence: web_search, db_query, and file_read have no timeout.
      Impact: Hung tool call can stall the event loop indefinitely.
      Required fix: Explicit timeouts on every tool call.

    Verdict: FAIL

    Connects securely to your tools. The creator never sees your data.

    About this skill

    This skill performs a principal-engineer review of agent system reliability and failure semantics.

    It examines system boundaries, state handling, tool failure modes, retry behavior, timeout design, recovery paths, resource bounds, and observability of failures. Every finding is classified by severity with evidence, impact, and a required fix. No praise, no filler — only actionable engineering findings.

    Use it when you need a hard review of reliability design before production, during incident analysis, or when auditing recovery behavior. The output is strictly severity-ranked and ends with a clear PASS / CONDITIONAL / FAIL verdict.

    Built for senior and principal engineers who require production-grade standards, not demo-level reviews.

    Does not suggest new tools or frameworks. Requires a detailed description of the system architecture to provide meaningful results.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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 today

    Permissions

    Read Files
    Listedtoday

    What's inside

    Frequently Asked Questions

    The skill identifies critical risks like infinite reflection loops, unhandled API timeouts, and state corruption that cause AI agents to fail silently or incur high costs in production.

    Yes, this skill is framework-agnostic; it audits the underlying logic and architecture whether you are using LangChain, CrewAI, AutoGen, or a custom-built orchestration layer.

    You will receive a detailed audit report featuring severity-ranked findings (Critical to Low), specific evidence-backed fixes, and a final reliability verdict of PASS, CONDITIONAL, or FAIL.

    You simply need to provide the system design, architecture diagrams, or the specific code blocks responsible for tool execution and error handling for the skill to begin its analysis.

    Unlike general prompts, this skill is strictly configured as a senior-level auditor that ignores surface-level praise to focus exclusively on technical edge cases and operational bounds.

    This skill focuses on the structural logic of your agent; while it doesn't execute live tests, it identifies the architectural flaws that lead to failures before you deploy to production.

    New in Agents & Orchestration