Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Webhook Delivery Reliability Audit

    1

    Audit webhook logs by event and destination. Separate retries and duplicate acknowledgements, flag signature failures, and measure delivery, timeliness and missing evidence with explicit denominators.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Webhook Delivery Reliability Audit

    Example session with this skill installed

    Synthetic demonstration request

    Audit orders.webhook deliveries for 2026-11-02, events emitted 10:00:00–10:07:00 UTC, one per minute. Expected events evt-1..evt-8 (source of truth is complete). SLA: a 2xx within 120 s of emit. Success = 2xx; trusted = 2xx AND signature verified. Evidence cutoff 10:15:00 UTC. Report coverage, delivery, timeliness, duplicates, signature failures and next investigations. Do not resend anything.

    Rows are delivery attempts, not a status stream. Times are UTC on 2026-11-02.

    RowEventAttemptEmitFinishStatusFinalSig OK
    1evt-1a110:00:0010:00:03200trueyes
    2evt-2a110:01:0010:01:02500false-
    3evt-2a210:01:0010:01:40200trueyes
    4evt-3a110:02:0010:02:04200trueno
    5evt-4a110:03:0010:03:03200trueyes
    6evt-4a210:03:0010:03:30200trueyes
    7evt-5a110:04:0010:04:05503false-
    8evt-5a210:04:0010:06:00503false-
    9evt-5a310:04:0010:09:00503true-
    10evt-6a110:05:0010:08:30200trueyes
    11evt-8a110:07:0010:07:02200trueyes

    evt-7 (emitted 10:06:00) has no attempt row.

    • Read your context and instructions
    • Compiled the webhook delivery reliability

    Six of eight expected events delivered a 2xx by cutoff; five were on time and five were signature-verified; four satisfied both conditions. evt-7 has no observation in a complete export, and evt-5 exhausted three attempts returning 503. evt-4 was delivered twice (idempotency risk) and evt-3 was accepted with an invalid signature (security finding).

    Contract: orders.webhook; 8 due events evt-1..evt-8; 120 s SLA; cutoff 10:15:00; complete export; signature required for trust. Eleven attempts cover seven events.

    MeasureResult
    Due events8
    Outcome reconciliation6 delivered + 1 failed + 1 missing = 8
    Observed coverage7/8 = 87.50%
    Delivery by cutoff6/8 = 75.00%
    On-time delivery (<=120s)5/8 = 62.50%
    Trusted delivery (2xx + sig)5/8 = 62.50%
    Retry recoveries1 (evt-2)
    Duplicate deliveries1 (evt-4) - idempotency risk
    Signature failures1 (evt-3) - security finding
    EventOutcomeEvidence / next check
    evt-2Delivered, on time via retryRows 2-3: 500 then 200 at 10:01:40 (40 s)
    evt-3Delivered, signature invalidRow 4: verify signing secret and verification path before trusting payload
    evt-4Delivered twiceRows 5-6: two 2xx for one event; confirm the receiver is idempotent
    evt-5Failed (exhausted)Rows 7-9: three 503s, final at 10:09:00; inspect upstream cause
    evt-6Delivered, lateRow 10: first 2xx at 10:08:30, 210 s > 120 s SLA
    evt-7MissingNo row in a complete export; compare dispatch history to ingestion

    First-success latencies: 3, 40, 4, 3, 210, 2 s -> min/median/max = 2/3.5/210 s (n=6). These are delivery latencies, not uptime. Investigate evt-7 and evt-3 first (unknown disposition, verification failure), then evt-4, evt-5, evt-6. No production changes were made.

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

    About this skill

    Stop trusting a wall of 200s

    A webhook can return 200 twice for the same event, succeed only on its third retry, or never arrive at all. Counting success lines hides every one of those. This skill guides your AI agent through a read-only audit of webhook delivery logs, keeping logical event–destination pairs separate from delivery attempts so retries and duplicates do not distort the numbers.

    What you receive

    • A SKILL.md workflow plus a synthetic worked example.
    • A clear audit contract: expected events, window, evidence cutoff, SLA, success and trust definitions.
    • Reconciled outcomes per expected event–destination pair: delivered, failed, unresolved, and missing or unknown evidence.
    • Observed coverage, delivery, on-time and trusted-delivery rates with explicit denominators.
    • Duplicate-delivery (idempotency) flags, signature-verification failures, retry recoveries, latency, and traceable next-step investigations.

    How to use it

    Ask your compatible agent to audit your webhook delivery logs. Provide a redacted export or pasted records (event id, endpoint, attempt, status, timestamps, final flag, signature result), the list of events that were expected, a time zone, a cutoff and an SLA. The agent produces the report; you decide what to investigate.

    Boundaries

    Instruction skill, not a hosted monitor or a webhook replayer. It does not connect to providers, resend deliveries, change endpoints, or guarantee delivery. No required paid API. An incomplete export cannot prove a missing delivery; it is reported as unknown coverage. Large exports need an analysis-capable agent with local tools.

    Validation

    Frontmatter and package validation pass. The worked-example arithmetic reconciles (6 delivered + 1 failed + 1 missing = 8 due events) and was checked programmatically. The demonstration is synthetic, not customer production data. Results depend on the quality of supplied evidence and the agent used.

    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.

    Read-only instruction skill, not monitoring, a replayer, or an integration. Cohort rates require expected event–destination pairs and usable correlation/time evidence; on-time metrics require an SLA. Incomplete exports cannot prove non-delivery, and HTTP acknowledgements do not prove downstream business processing. Signature trust requires explicit evidence. Ordering needs source sequence. Large exports require local analysis tools; results depend on evidence and agent capability.

    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

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 9 days ago

    Permissions

    Read Files

    Read-only analysis of user-provided webhook logs. File reading is needed only for supplied files; terminal/local analysis is optional for larger exports. No network, credentials, environment variables, provider connections, or production writes are required.

    Listed9 days ago

    What's inside

    Frequently Asked Questions

    It groups attempts by source, event and destination, so retries and duplicate rows do not inflate success or hide a failed destination. Coverage, delivery, timeliness and verified delivery use explicit denominators.

    Pasted records or an export your agent can read, with event id, endpoint, attempt id, status, timestamps, a final flag and a signature result. Provide the list of expected events so a missing delivery can be distinguished from unknown coverage.

    No. It is a text instruction skill and a Markdown reference. You need a compatible agent and your own log data. Your agent provider may have its own usage costs.

    An explicit signature-verification failure on an accepted delivery warrants investigation. Missing verification evidence is reported as unknown, not invalid. Neither a 2xx nor a signature alone proves downstream business processing.

    No. Even a complete export only establishes a missing observation within its scope. With an incomplete export, missing rows are unknown coverage; dispatch and receiver evidence are needed to establish what happened.

    No. It analyzes supplied evidence and recommends investigation steps. It does not connect to providers, replay deliveries, change endpoints, or run anything.

    New in Testing & Debugging