Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Cycling HR Thresholds & Zones from Garmin, Strava or CPET

    2

    Turns your ride history (and a lab CPET if you have one) into a defensible cycling LTHR, VT1 estimate, and Garmin-ready heart-rate zones, with the evidence and uncertainty shown.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Cycling HR Thresholds & Zones from Garmin, Strava or CPET

    Cycling HR Thresholds & Zones from Garmin, Strava or CPET

    Example session with this skill installed

    I've been racing crits and road races this season on Garmin, no power meter. Use my last three months of rides to work out my cycling LTHR and set up zones I can type into Garmin. Tell me what you can and can't conclude.

    • Read your context and instructions
    • Compiled the cycling hr thresholds

    Executive result

    ConstructEstimateRangeConfidenceSource
    Cycling LTHR / LT2186 bpm184–189ModerateField, repeated 20-min race windows
    VT1 / GETNot identifiable from HR aloneprovisional ceiling ≤158LowField, one 60-min steady ride
    VT2 / RCPNot supplied——No CPET provided
    HRmax (observed)204 bpm—Observed, not testedSprint session

    Values were not averaged across sources. LTHR comes from clustering of sustained windows; VT1 is a preview bound only.

    Field audit

    • Source: Garmin, 2026-07-20 to 2026-10-20, 73 cycling activities, pagination complete. No synchronized Strava copies in scope.
    • Six rides inspected on raw 1 Hz records (two races, two hard sessions, one long climb ride, one steady 60-min ride). HR coverage 100% on all six. Power: none. Cadence present on four. HR sensor type not exposed in the FIT device table, so treated as unknown.
    • Best continuous windows (bpm, no gap bridging over 5 s):
    RideDurationTemp20 min30 min40 minNotes
    Road race A177 min32 °C188.4180.1179.0surges to 202; HR 1st half 183 → 2nd half 173 (heat + fatigue drift)
    Criterium B36 min16 °C188.2——nearly flat 186–190 for 30 min, then sit-up
    Climb ride C183 min36 °C164.2158.7—not a threshold effort
    Hard session D63 min30 °C———1 Hz but efforts fragmented by stops
    Steady 60-min E86 minn/a———3 s recording, 5-min blocks 153–164, mean 159, drift 160 → 158
    • Interpretation: the two race 20-minute windows agree at 188 bpm in opposite temperatures (16 °C vs 32 °C), which argues the number is physiological rather than heat-driven. Race windows include surges above threshold, so the operational LTHR is set slightly below the 20-minute mean:

    186 bpm, range 184–189. The 30/40-minute windows of the road race (179–180) are depressed by mid-race recovery and heat, not evidence of a lower threshold.

    • VT1: without power, the 60-minute steady ride cannot separate workload from drift. Its negligible drift at a 159 mean suggests VT1 is at or above the high 150s but cannot pin it. Reported as

    not identifiable; 158 is used only as a provisional ceiling for the zone preview below.

    CPET audit: not supplied.

    Reconciliation: single source (field). LTHR preferred from repeated race windows; VT1 left open. Nothing averaged.

    Zones (dual-threshold five-zone model, preview with provisional VT1 = 158, LTHR = 186, HRmax = 204)

    ZoneBPM%LTHRDomain
    Z1open–142open–76.3%Domain 1 (lower)
    Z2143–15776.9%–84.4%Domain 1 (upper)
    Z3158–16784.9%–89.8%Domain 2 (lower)
    Z4168–18590.3%–99.5%Domain 2 (upper)
    Z5186–204100.0%–109.7%Domain 3

    Device setup (proposed, not yet applied): Garmin custom lower boundaries Z2=143, Z3=158, Z4=168, Z5=186; LTHR 186 in the athlete profile. If Garmin forces its fixed template, its Z2 would read 136–164 bpm, which overlaps the provisional VT1; keep custom bpm zones.

    Caveats: no power, so workload and cardiac drift are confounded; sensor modality unknown; VT1 is a preview bound, not a measurement; the Z2/Z3 edge inherits that uncertainty.

    Best added test: one staged ramp on a calibrated trainer at your normal cadence with a chest strap and either breath-by-breath gas exchange or end-stage lactate. It resolves VT1, confirms LTHR, and gives you a real HRmax in a single session.

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

    About this skill

    What it does

    Point your agent at your Garmin or Strava history and it will find the rides that actually say something about your threshold, measure the best sustained 20/30/40/60-minute heart-rate windows on the raw time series, check each candidate for heat, drift and sensor gaps, and give you a cycling LTHR with a range and a confidence level. If you have a breath-by-breath CPET export, it detects VT1/GET and VT2/RCP candidates with the standard gas-exchange markers, aligns the metabolic cart and ergometer clocks, and reconciles lab and field results construct by construct instead of averaging them. It finishes with a dual-threshold five-zone table in bpm and %LTHR, the exact lower boundaries to type into Garmin, and the single test that would most reduce the remaining uncertainty.

    Why this instead of a prompt

    Most "calculate my zones" prompts take your Garmin threshold at face value, convert 20-minute max HR with a fixed percentage, and silently treat a running threshold as a cycling one. This skill refuses those shortcuts. It works on raw records, not summary numbers. It treats configured Garmin values as metadata to verify, not evidence. It never averages a lab VT1 with a field LTHR. And it reports what it could not identify from your data rather than inventing a number.

    Key features

    • Best-window analysis on raw FIT/HR records with explicit gap handling (no bridging of pauses).
    • Independent field and laboratory pipelines, then a reconciliation table by physiological construct.
    • CPET analysis with V-slope, VE/VO2, VE/VCO2, PetO2/PetCO2 markers and multi-window smoothing checks.
    • FTP proxies when you have no power meter, each labeled by method (lab ergometer, climb physics, critical power).
    • Dual-threshold five-zone model in bpm and %LTHR, plus Garmin custom-zone entries and the fixed-percentage mismatch.
    • Capability discovery: uses whichever of Garmin MCP, the official Strava MCP, or local files is available in the session.
    • Tested Python helpers (unit tests included) and a report contract so every run has the same sections.

    Who it is for

    Cyclists and triathletes training by heart rate, coaches building zones for athletes without power, and anyone holding a CPET report who wants it turned into usable training boundaries.

    How it is built

    One SKILL.md workflow, five reference documents (field workflow, CPET methods, reconciliation and zones, client routing, report contract), four Python CLIs with tests, and an agents/openai.yaml for Codex. Runs on Python 3.10+ with numpy and pandas.

    Known limitations

    • Heart rate alone cannot locate VT1 confidently. Without controlled workload, power, lactate or gas exchange the skill reports VT1 as "not identifiable" and only offers a provisional ceiling for a zone preview.
    • Garmin access needs a working Garmin MCP (for example Taxuspt/garmin_mcp) or exported FIT/JSON files. Strava access needs the official Strava MCP, which requires a Strava subscription and exposes summary data only.
    • Optical wrist HR is usable but flagged; chest-strap data is preferred for threshold work.
    • Not medical advice. It does not diagnose conditions and does not publish names, coordinates or activity IDs.

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

    Heart rate alone cannot locate VT1 confidently; without power, lactate or gas exchange it reports VT1 as not identifiable. Garmin needs a working Garmin MCP or exported FIT/JSON; Strava needs the official Strava MCP (subscription, summary data only). Not medical advice.

    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 27 days ago

    Software
    Python 3.10+NumPypandasopenpyxl
    Garmin MCP
    orFIT/JSON files
    Issuer MCP
    Optional
    Strava MCPStrava subscriptionCPET export (CSV or XLSX)

    Install the Python packages from scripts/requirements.txt. Taxuspt/garmin_mcp works as the Garmin MCP. The official Strava MCP needs a Strava subscription.

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    Allowed Hosts

    mcp.strava.com
    www.strava.com
    connect.garmin.com

    Shell and file access run the bundled Python helpers on the user's own exported activity or CPET files. Network access is only through the Garmin or Strava MCP the user has already connected; the skill makes no calls of its own and never uploads health data anywhere.

    Listed27 days ago

    What's inside

    Frequently Asked Questions

    No. Without power the skill still estimates LTHR from sustained HR windows and labels every FTP proxy by method. It tells you which conclusions power would unlock.

    No. It proposes custom zone boundaries and shows a preview. You enter them, or ask your agent to, as a separate step.

    Because heart rate alone cannot locate the first threshold unless workload was controlled (power, lactate or gas exchange). The skill would rather say so than invent a number. Give it a CPET export and it will identify VT1 and RCP candidates with the standard markers.

    You will receive the structured AI agent configuration, the specialized Python logic for threshold detection, and a `requirements.txt` file for the necessary data processing libraries.

    No, this skill is designed for performance analysis and training zone optimization; it is not a medical device and should not be used to diagnose any cardiovascular or respiratory conditions.

    New in Research & Analysis