Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Scheduled Job Reliability Audit

    1

    Audit recurring-job logs without inflating success rates. Separate retries, late completions, missing observations and unfinished work, then produce traceable metrics and a prioritized investigation plan.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Scheduled Job Reliability Audit

    Example session with this skill installed

    Synthetic demonstration request

    Audit hourly invoice-export runs for 2026-11-01 from 00:00 through 08:00 UTC, excluding 08:00. Each slot must finish within 30 minutes. The evidence cutoff is 08:00 UTC. This export is complete for the window, and final=true means no further retries are pending. Report delivery, timeliness, missing evidence and next investigations. Do not rerun anything.

    All timestamps below are UTC on 2026-11-01. These are attempt snapshots, not a stream of status-change events.

    RowSlotAttemptStartFinishStatusFinalError
    100:00a000:0000:10successtrue
    201:00a101:0001:05failedfalseupstream timeout
    301:00a1r01:0601:15successtrue
    402:00a202:0002:10failedtrueexport rejected
    504:00a404:0004:40successtrue
    605:00a505:00runningfalse
    706:00a606:0006:10successtrue
    806:00a606:0006:10successtrue
    907:00a707:0007:05cancelledtrueoperator cancellation
    • Read your context and instructions
    • Compiled the scheduled job reliability

    Four of eight due slots have delivered successfully by cutoff; three were on time. The 03:00 slot has no observation in the complete export, and the 05:00 attempt remains running 150 minutes past its deadline. The retry at 01:00 recovered within its deadline and does not constitute a missed delivery.

    Contract: job invoice-export; expected slots 00:00–07:00 UTC; 30-minute deadline; cutoff 08:00 UTC; complete export; explicit final-state flags. Row 8 duplicates row 7 and is excluded. There are eight distinct attempts across seven observed slots.

    MeasureResult
    Due slots8
    Outcome reconciliation4 success + 1 failed + 1 cancelled + 1 unresolved + 1 missing = 8
    Observed slot coverage7/8 = 87.50%
    Delivery by cutoff4/8 = 50.00%
    On-time delivery3/8 = 37.50%
    Closed-slot success4/6 = 66.67%; excludes the missing and unresolved slots
    Retry recoveries1: 01:00
    Optional terminal-attempt success4/6 = 66.67%; excludes one cancellation and one running attempt
    SlotOutcomeEvidence / next check
    01:00Success, on timeRows 2–3: timeout followed by successful retry, finished 01:15
    02:00FailedRow 4: terminal rejection; inspect the rejection details before proposing a fix
    03:00Missing observationNo row; compare scheduler dispatch history and export ingestion coverage
    04:00Success, lateRow 5: first success at 04:40, 10 minutes beyond its deadline
    05:00UnresolvedRow 6: still running at cutoff; inspect worker heartbeat/queue state before any rerun
    07:00CancelledRow 9: cancellation is recorded; confirm whether it reflects intended skipped work

    First-success latencies are 10, 15, 40 and 10 minutes; min/median/max are 10/12.5/40 minutes (n=4). These are job-delivery measurements, not availability or uptime.

    Investigate the 03:00 evidence gap and overdue 05:00 execution first because their disposition is unknown. Then inspect the terminal 02:00 rejection and the 04:00 execution trace. Confirm the 07:00 cancellation policy. The sample does not establish an infrastructure outage or explain the late completion's cause. All steps above are investigation recommendations; no production changes were made.

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

    About this skill

    Find the delivery failures hidden by retry counts

    A job can fail twice and still deliver on time. Another can disappear from your logs entirely. This skill guides your AI agent through a read-only audit of scheduled work, keeping expected deliveries separate from execution attempts.

    What you receive

    • A SKILL.md workflow plus two synthetic worked examples.
    • A clear audit contract: time window, evidence cutoff, expected schedule and completion deadline.
    • Reconciled counts for success, failure, cancellation, unresolved work and missing or unknown evidence.
    • Observed coverage, delivery and on-time rates with explicit denominators.
    • Retry recovery and latency analysis, with traceable source rows and prioritized investigation steps.

    How to use it

    Ask your compatible agent to audit your recurring job logs. Provide a redacted export or pasted records, expected slots or an unambiguous schedule, time zone, cutoff and deadline. The agent produces the report; you decide which investigations or changes to perform.

    Boundaries

    This is an instruction skill, not a hosted monitoring service or executable scheduler integration. It does not connect accounts, rerun jobs, repair systems, or guarantee uptime. It has no required paid API. Large exports need an analysis-capable agent with suitable local tools. Unknown export completeness is reported explicitly; missing log rows are not automatically called missed executions.

    Validation

    Frontmatter and package validation passed. Worked-example arithmetic was checked with Python. An independent agent applied the instructions to a separate synthetic case covering a queued retry, an exact duplicate, a missing slot, and an early completion. The demonstration below is synthetic, not customer production data. Results depend on the quality of supplied evidence and the agent used.

    Read-only instruction skill, not continuous monitoring or an executable integration. Requires usable logs, correlation keys, an expected schedule, a time zone and deadlines for slot-level metrics. Incomplete exports cannot prove missed executions. Ambiguous cron/DST schedules need confirmed expected slots. Large exports require local analysis tools. Outputs depend on supplied evidence and the agent used.

    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 9 days ago

    Permissions

    Read Files

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

    Listed9 days ago

    What's inside

    Frequently Asked Questions

    It separates scheduled deliveries from execution attempts, so retries and duplicate records do not inflate slot success. It reports coverage, delivery and timeliness with explicit denominators.

    Use pasted records or an export your agent can read, with job identity, scheduled slot, attempt identity, status and timestamps. No native scheduler connectors are bundled. An explicit expected-slot list is safest for ambiguous schedules.

    No. The product 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.

    One SKILL.md workflow and one reference containing two synthetic worked examples. The workflow covers normalization, slot outcomes, metric definitions, evidence limits and investigation reporting. There is no executable metric engine or monitoring service.

    No. It analyzes supplied evidence and recommends investigation steps. It does not connect to schedulers, execute commands from logs, change schedules or rerun jobs.

    No. With a confirmed complete export, it identifies a missing observation. With an incomplete export, it reports unknown coverage. Scheduler history or other evidence is needed to establish what actually happened.

    New in Testing & Debugging