Triage a failed Claude Code turn

    1

    Diagnose a failed or slow turn from the session log: upstream quota, relay fault, transport, or a retry that hid itself.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Triage a failed Claude Code turn

    Example session with this skill installed

    A request failed with an empty error message, then succeeded when I retried. Was it me or the provider?

    • Read your context and instructions
    • Compiled the triage a failed

    Summary

    Mechanism: upstream quota (HTTP 429). The client retried twice and recovered on the third attempt.

    FieldValue
    Attempts3
    Final status200 (recovered)
    Upstream error429 rate_limit_exceeded
    Wall time41.2s, of which 38.6s was retry backoff

    Why it looked like a hang

    The retries are not shown in the UI. A turn retried three times looks identical to a clean turn — it just takes longer.

    Ruled out

    • Not auth: the key resolved and the request reached the provider.
    • Not transport: no connection reset or TLS failure in the log.
    • Not a client bug: the error text came from upstream, verbatim.

    Read-only. Nothing was retried or modified by this skill.

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

    About this skill

    When a Claude Code request fails, the UI often says very little. The log says what actually happened: the provider's own error text, the HTTP status, and how many times the client retried before giving up.

    That last part is the useful one. A turn that was retried three times and then succeeded looks identical to a clean turn in the UI - it just takes longer. Without the log you cannot tell a slow model from a flaky upstream.

    This skill names the mechanism (quota, auth, relay, transport, timeout) and quotes the evidence, rather than guessing from the symptom.

    It also separates two things that look alike: an error that was recovered (latency cost) and a failure that ended the turn (work lost).

    Read-only. Node 20+.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.

    Node.js 20+ required. Read-only access; does not modify logs or retry failed turns. Cannot infer internal relay routing logic not present in logs.

    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 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 2 days ago

    • Free with an Agensi account

    Software
    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    This skill provides a deep-dive analysis of Claude Code session logs to distinguish between network timeouts, API rate limits (TPM/RPM), and account-level credit issues that are often obscured in the standard CLI output.

    The skill requires Node.js 20+ and is specifically designed to work with Claude Code session logs; it functions as a standalone analyzer that processes local log files to generate diagnostic JSON reports.

    Yes, the skill includes a '--since' flag that allows you to isolate and analyze specific timeframes, making it easy to debug a single slow turn or a specific failure event from earlier in the day.

    You will receive the 'cf-analyze.mjs' utility, the core triage logic library, and documentation on how to interpret the four primary fault classifications and telemetry data.

    The tool identifies 'hidden retries' where a request initially fails but eventually succeeds; it audits the retry count and latency per attempt to show you exactly why a response took longer than expected.

    New in Testing & Debugging

    Free