Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    JMeter Performance Analyzer

    1

    Analyze Apache JMeter CSV/XML JTL results and generate deterministic performance metrics, optional SLO evaluations, and JSON/Markdown reports.

    $7

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    JMeter Performance Analyzer

    JMeter Performance Analyzer

    Example session with this skill installed

    Authentic local OpenCode acceptance example

    Selected input: ten-sample.csv.jtl (10 synthetic raw CSV samples, 451 bytes). This demonstrates local execution, not Agensi-delivered execution.

    The following is a verified verbatim excerpt from the historical request template; the original temporary path and permission-harness instructions are omitted. It is not the complete original prompt.

    The format is CSV. Export timestamps are sample STARTS in epoch milliseconds.
    No SLOs are supplied. Do not infer a threshold or performance pass/fail.

    Use the actual installed helper, retain canonical JSON and deliver its generated human-readable
    Markdown UNCHANGED, in full, with no preface, wrapper or appended commentary. Preserve the
    fenced label blocks, exact fractions, limitations and final newline if the transport supports it.
    Do not invent measurements, causes, utilization, capacity, anomalies or regressions.

    • Read your context and instructions
    • Compiled the jmeter performance analyzer

    JMeter Performance Report

    Executive summary

    • Analysis state: complete. Offline saved-sample analysis; no load test was executed.
    • Population: 10 samples; 9 successful; 1 failed. Failures are included in latency statistics.
    • Error rate: 10.0 percent.
    • Latency p90 / p95 / p99 (ms): 190.0 / 595.0 / 919.0.
    • Throughput: ~1.1049723756906078 (exact 200/181) samples/s.
    • No supplied SLOs; no performance pass/fail is assigned.

    Technical evidence

    • Input: csv-raw-v0.1; 451 original bytes; validation valid.
    • Original-byte SHA-256: bfa04d8dc1504e78db268d8677cdccad54b1e0fd86275a91f4797e56f1547ac1.
    • Helper: 0.1.0-report.1; ledger schema: 0.1.0.
    • Population: accepted raw samples; XML uses leaves only. Labels are sampler identities, not necessarily endpoints.
    • Percentiles: linear interpolation at (p / 100) * (n - 1), using the full accepted population.
    • Latency min / mean / median / max (ms): 50 / 180.0 / 100.0 / 1000.
    • Timestamp semantics: start.
    • Observation window (UTC epoch ms): 1700000000000 through 1700000009050; duration 9050 ms.
    • Overall and per-label throughput use the same run window. A '~' view is approximate; its exact fraction is authoritative.

    Per-label observations

    Order: failed count descending, then exact error rate, exact p95, and exact label text. This is relative ranking, not an SLO verdict.

    Label referenceSamplesFailedError percentMean msp95 msp99 msSamples/s
    L0034125.0325.0865.0973.0~0.4419889502762431 (exact 80/181)
    L001400.0100.0100.0100.0~0.4419889502762431 (exact 80/181)
    L002200.050.050.050.0~0.22099447513812154 (exact 40/181)

    Label identities

    Labels below are inert JSON string literals; escapes preserve identity.

    L001
    "GET /catalog"
    
    L002
    "GET /health"
    
    L003
    "POST /checkout"
    

    Supplied SLO outcomes

    None supplied. No customer threshold, baseline or compliance result is invented.

    Evidence-based findings and prioritized next checks

    • Observed failures: overall.failure_count is 1; L003 recorded 1 of 4 samples as failed. Exported success flags, not response-code assumptions, determine these counts.
    • Exploratory ordering: L003 has the highest observed p95, 865.0 ms, over 4 samples. This relative position is not an anomaly, regression or threshold breach by itself.

    Priority below is a transparent triage ordering, not an estimate of severity, effort or guaranteed benefit.

    1. Start failure triage with L003, then the remaining ranked labels; correlate sample/test evidence without assuming a cause.
    2. Review workload mix and sample size for L003; obtain discriminating traces if further latency investigation is warranted.

    Limitations

    • Saved JTL evidence does not establish CPU, memory, database or network utilization, root cause, concurrent users or system capacity. No causal hypothesis is confirmed.
    • No cross-run baseline comparison, performance-regression test, temporal bins or anomaly detection was performed.
    • Small-population tail estimates are descriptive; relative rankings and recorded sample outcomes are not general workload guarantees.
    • Structural validation accounts for this selected export; externally missing samples or well-formed truncation cannot always be detected.
    • Runtime resource targets and marketplace compatibility are not established by this report. Model-provider connectivity is separate from offline helper analysis.

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

    About this skill

    Turn saved JMeter results into a clear performance report

    JMeter Performance Analyzer helps performance QA engineers, QA automation engineers, SDETs and backend teams investigate one saved, sanitized raw-sample JMeter JTL export. Get deterministic calculations and a structured report without rerunning a load test or asking a language model to calculate the statistics.

    What you get

    • Supported raw CSV and XML JTL analysis: validates the selected export and reports unsupported or malformed input explicitly.
    • Overall and per-label latency: sample counts, minimum, mean, median, maximum, p90, p95 and p99. Percentiles use documented linear interpolation, and recorded failed samples remain included.
    • Failure and error reporting: successful/failed sample counts and error percentages based on recorded sample success flags.
    • Common-window throughput: overall and per-label rates share the observed run window. Missing or unknown time information produces an unavailable result, not a made-up rate.
    • Optional supplied SLO evaluation: compare explicitly supplied thresholds using exact values, with distinct met, not_met and not_evaluated outcomes. No default customer thresholds are invented.
    • JSON and Markdown deliverables: a canonical numeric ledger and a readable executive/technical report with input SHA-256 provenance, per-label observations, supplied SLO outcomes, limitations and prioritized investigation suggestions.

    How to use it

    1. Make the delivered skill instructions, README and standalone Python helper accessible through your agent's supported delivery route.
    2. Provide one sanitized local raw CSV or XML JTL file and identify its format. Specify whether timestamps record sample starts or ends when known; unknown semantics leave throughput unavailable.
    3. Ask for the Markdown report or canonical JSON. Add explicit SLO definitions only when you want threshold evaluation.

    The agent invokes the supplied helper and delivers its generated output. For Markdown, it retains the JSON ledger and verifies matching input/helper identities before delivering the report. Results stay in the interaction unless you separately authorize saving them.

    Requirements

    Python 3.12 or later, permission to run the delivered helper, access to the selected JTL, and complete stdout/stderr/exit-status capture are required. The helper uses the Python standard library; no JMeter installation, Java, MCP server, external Python package, service credential or separate model API key is required by the analysis workflow. Your agent provider's normal connectivity and billing remain separate.

    Scope and limits

    This is offline analysis of one supported raw-sample export, not a load-test runner. It does not open JMX plans, probe services, ingest aggregate reports, compare runs, detect temporal anomalies, establish infrastructure root cause or prove system capacity. Input limits include 10 MiB, 50,000 raw samples and 200 labels, with additional structural bounds documented in the included README. JSON is capped at 2 MiB and Markdown at 512 KiB; the helper reports limit errors rather than silently truncating results. Host capture limits may be smaller.

    Local functional/native acceptance and package verification have passed for the reviewed development baseline. P02 delivery and helper execution through Agensi have not yet been demonstrated; the receiving host must support the requirements above. No universal host/OS compatibility or processing-time guarantee is claimed.

    Creator and license

    Created by Ramiro Maceda — Senior Software Quality Assurance Engineer.

    Repository: https://github.com/rmaceda7-coder/IAagentSkills

    The package includes canonical creator details in AUTHORS.md and the proprietary notice in LICENSE.md. Copyright © 2026 Ramiro Maceda in his original skill materials; third-party materials and trademarks remain with their respective owners. Buyer rights and restrictions are governed by the applicable Agensi Terms of Service as referenced in LICENSE.md; this listing adds no buyer rights or off-platform distribution permission. Naming JMeter describes the supported result format, not an endorsement.

    • One supported raw-sample CSV/XML JTL per analysis; no aggregate/weighted reports, JMX execution, traffic generation, live monitoring or hosted dashboards.
    • Input quality and recorded success/timestamp fields determine the results. CSV/XML must match the documented profiles; unsupported or malformed input rejects.
    • Throughput requires sufficient timestamps, explicit start/end semantics and a nonzero observation window; otherwise it remains unavailable.
    • Input ceilings: 10 MiB, 50,000 raw samples and 200 labels, plus documented structural bounds. Output ceilings: 2 MiB JSON and 512 KiB Markdown; no silent truncation. Host limits may be smaller.
    • SLO evaluation uses only explicitly supplied supported definitions. Invalid/duplicate/unevaluable definitions can remain not_evaluated; analysis completion is not a performance pass.
    • No cross-run regression comparison, temporal anomaly detection, confirmed infrastructure cause, concurrency measurement or capacity guarantee.
    • Sanitized input, Python execution and complete capture are required. Runtime time/memory targets and universal host/Agensi compatibility are not guaranteed.

    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 7 days ago

    Requires

    Python 3.12 or later (standard library only)
    A sanitized local raw Apache JMeter CSV or XML JTL file
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Read only the user-selected sanitized JTL and delivered skill/support/helper files. Execute the unchanged Python helper with literal arguments and capture stdout/stderr/exit status. Temporary helper writing/removal is needed for Agensi delivery. Reports stay in the interaction by default; writing to a selected output directory requires explicit user authorization. No helper network access, live load test, service credentials or arbitrary file discovery is required. The host model may use its normal provider connection. Permission declarations are not a verified sandbox profile.

    Listed7 days ago

    What's inside

    Frequently Asked Questions

    No. It analyzes one saved supported raw-sample CSV/XML JTL using a local Python helper. It does not execute JMX plans, generate traffic, probe services, provide live monitoring or modify a target system.

    Python 3.12+, the delivered skill/support text and standalone helper, a sanitized supported local JTL, permitted local execution, and complete output/status capture. The helper uses only the Python standard library; no JMeter, Java, MCP server or external Python packages are required.

    One supported raw-sample CSV or XML JTL per analysis. CSV must be headered UTF-8/comma-separated with the required exact fields. XML must use the documented testResults/sample profile. Aggregate/weighted, malformed and unsupported profiles reject rather than being silently repaired.

    Canonical JSON or a generated executive/technical Markdown report with overall/per-label metrics, input SHA-256 provenance, supplied SLO outcomes and limitations. For Markdown the agent retains JSON and checks input/helper identities. Output stays in the interaction by default; saving to your selected output directory requires explicit authorization.

    Yes. Supply explicit supported definitions with metric, scope, operator, decimal-string threshold and units. Exact values govern met/not_met verdicts; invalid, duplicate or unavailable comparisons can remain not_evaluated. No threshold is invented and analysis completion is not an overall performance pass.

    Created by Ramiro Maceda, Senior Software Quality Assurance Engineer. AUTHORS.md contains canonical attribution and LICENSE.md contains the approved proprietary notice. Buyer rights and restrictions are governed by the applicable Agensi Terms of Service referenced there; no additional buyer or off-platform distribution rights are granted.

    New in Testing & Debugging