Pp Garmin

<!-- GENERATED FILE — DO NOT EDIT.
     This file is a verbatim mirror of library/health/garmin/SKILL.md,
     regenerated post-merge by tools/generate-skills/. Hand-edits here are
     silently overwritten on the next regen. Edit the library/ source instead.
     See the repository agent guide, section "Generated artifacts: registry.json, cli-skills/". -->

Garmin Connect — Printing Press CLI

Prerequisites: Install the CLI

This skill drives the garmin-pp-cli binary. You must verify the CLI is installed before invoking any command from this skill. If it is missing, install it first:

1. Install via the Printing Press installer. It defaults binaries to $HOME/.local/bin on macOS/Linux and %LOCALAPPDATA%\Programs\PrintingPress\bin on Windows:

   npx -y @mvanhorn/printing-press-library install garmin --cli-only

2. Verify: garmin-pp-cli --version
3. Ensure the reported install directory is on $PATH for the agent/runtime that will invoke this skill.

If the npx install fails (no Node, offline, etc.), fall back to a direct Go install (requires Go 1.26.6 or newer). This installs into $GOPATH/bin (default $HOME/go/bin), so add that directory to $PATH instead:

go install github.com/mvanhorn/printing-press-library/library/health/garmin/cmd/garmin-pp-cli@latest

If --version reports "command not found" after install, the runtime cannot see the binary directory on $PATH. Do not proceed with skill commands until verification succeeds.

Garmin Connect caps every daily-stats request at 28 calendar days and issues no personal API token, which is why its ecosystem is a handful of Python libraries rather than a tool an agent can call. history walks the whole account into a local SQLite archive, oldest day first, and insights sleep and insights training answer months-long sleep, training-load and heart-rate-zone questions from that archive with JSON on stdout. auth login keeps each household account in its own home and refuses to store a token that belongs to somebody else.

When to Use This CLI

Reach for this CLI for questions about the owner's own Garmin Connect history: sleep and training trends, VO2 max, resting heart rate, step and intensity-minute totals, heart-rate zones, and the activity feed. It earns its keep on questions that span weeks or years, because those are answered from the local archive instead of from Garmin's 28-day pages. Do not reach for it before the archive is filled: an unfilled archive answers every trend question with nulls and a hint naming history, which is an honest answer but not the one the user asked for.

Anti-triggers

Do not use this CLI for:

• Writing anything back to Garmin Connect. Every command here reads.
• Another household member's account from this home. Each account needs its own login in its own home.
• A live device readout. A metric appears here only after the watch or bike computer has synced to Garmin Connect.
• Strava, Apple Health, Whoop or Oura data, which live in their own services.

Unique Capabilities

These capabilities aren't available in any other tool for this API.

Questions answered from the archive

• insights sleep — Duration, score, stage split and resting heart rate for a window of nights, with the equally long window before it beside them.

  Pick this over per-night endpoint calls whenever the question spans more than a few nights: one local read replaces one request per night, and nights the watch did not measure are counted as missing rather than as zero.

  garmin-pp-cli insights sleep --days 90 --agent --select period,coverage.nights_with_stats,duration_minutes.avg,score.avg

• insights training — Archived activities grouped by type with their time and distance, time summed per heart-rate zone, weekly load buckets, and the VO2-max and readiness trends beside them.

  Pick this over the activity feed when the question is about volume, intensity or balance rather than one workout; it reports how many activities had zones archived so a strapless ride is not read as easy training.

  garmin-pp-cli insights training --days 28 --agent --select totals,by_type.type,by_type.duration_minutes,hr_zones

Local state that compounds

• history — Fills a local SQLite archive of every Garmin daily series from the oldest day forward, keeping one bookmark per series so an interrupted run resumes instead of restarting.

  Run this before any question that reaches past the last 28 days: --since all or --since YYYY-MM-DD is the only depth control it offers, --status reports what each series still owes, --strict makes a scheduled run exit non-zero instead of warning past a series that did not answer, and every insights command reads the archive rather than the API.

  garmin-pp-cli history --since all

• sql — Ask the archive anything in SQL: one read-only SELECT across every series history has filled.

  Reach for this when the question is neither the sleep nor the training recipe: one SELECT joins any two series on the id they share, and a single-statement gate over a mode=ro, query_only handle means no query can write, whatever it says.

  garmin-pp-cli sql "SELECT substr(json_extract(data,'$.startTimeLocal'),1,7) AS month, count(*) AS activities, round(sum(json_extract(data,'$.duration'))/3600.0,1) AS hours FROM resources WHERE resource_type='activities' GROUP BY month ORDER BY month DESC LIMIT 12" --agent

Auth you can trust with a household

• auth login — Refuses to store a token whose account is not the address you named, so a shared browser cannot sign the wrong household member in.

  Run this once per Garmin account per home before anything else; afterwards every command works unattended from the refresh token.

  garmin-pp-cli auth login --email you@example.com

Command Reference

account — Bootstrap and identity: social profile, unit settings, and the account email a login is checked against.

• garmin-pp-cli account personal-information — Returns the identity block for the signed-in account, including the account email.
• garmin-pp-cli account settings — Returns the account's settings.
• garmin-pp-cli account social-profile — Returns the account's public profile.

activities — The activity feed, per-activity detail and splits, lifetime breakdowns, and original file downloads.

• garmin-pp-cli activities breakdown — Server-side aggregation of every activity on the account into totals per parent activity type, by duration or distance.
• garmin-pp-cli activities download-original — Fetches the original recorded file for one activity as a ZIP containing the device FIT file.
• garmin-pp-cli activities get — Everything Garmin holds about one activity at summary level: type, timing, distance and elevation.
• garmin-pp-cli activities hr-time-in-zones — Seconds spent in each configured heart-rate zone during one activity.
• garmin-pp-cli activities list — The account's activity feed, one row per recorded activity, newest first.
• garmin-pp-cli activities splits — The lap or split breakdown of a single activity, each with its own distance, duration and averages.

fitness — Fitness level over time: VO2 max, max-met values, and Garmin's fitness-age estimate.

• garmin-pp-cli fitness age — Garmin's fitness-age estimate and the components it was computed from for a single date.
• garmin-pp-cli fitness max-metrics — Fitness-level trend: VO2 max, generic and cycling, plus Garmin's max-met value, one row per date.

heart_rate — Configured heart-rate zones, daily heart-rate detail, and per-activity time in zone.

• garmin-pp-cli heart-rate daily — The intraday heart-rate series for a single date plus the resting, minimum and maximum values Garmin derived from it.
• garmin-pp-cli heart-rate daily-alt — Identical payload to daily, on the wellness-service path that takes the date as a query parameter.
• garmin-pp-cli heart-rate zones — The zone boundaries the account has configured, per sport.

sleep — Sleep score trends, server-aggregated nightly summaries, and full per-night detail.

• garmin-pp-cli sleep night — Everything Garmin recorded for a single night: stage minutes, sleep windows, restlessness and the sleep-score breakdown.
• garmin-pp-cli sleep night-alt — Identical payload to night, on the sleep-service path that takes the date as a query parameter.
• garmin-pp-cli sleep score-stats — The sleep score trend: one row per calendar date with the score value and qualifier.
• garmin-pp-cli sleep stats — One row per night between start and end, aggregated by Garmin.

steps — Daily and weekly step totals against the account's goal.

• garmin-pp-cli steps daily — One row per calendar date with total steps, the step goal and the distance walked.
• garmin-pp-cli steps weekly — Weekly step buckets instead of daily ones, about a year of them in a single request.

training — Training status over a window and the daily training-readiness score.

• garmin-pp-cli training readiness — Garmin's training-readiness score for a single date, with the inputs it was built from.
• garmin-pp-cli training status — Training status with acute and chronic load, for the days ending on the given date.

wellness — Daily wellness roll-ups: the day summary, intensity minutes, resting heart rate, and hydration.

• garmin-pp-cli wellness daily-summary — The single-day roll-up: steps, floors, intensity minutes, calories, resting heart rate and stress.
• garmin-pp-cli wellness hydration — Water intake logged for a single date against the day's goal.
• garmin-pp-cli wellness hydration-alt — The hydration payload plus the activity sweat-loss and goal-adjustment fields.
• garmin-pp-cli wellness intensity-minutes-weekly — Weekly buckets of moderate and vigorous intensity minutes against the account's weekly goal.
• garmin-pp-cli wellness metrics-daily — A single named metric as a daily time series over a date range, selected by the numeric metric id.

Finding the right command

When you know what you want to do but not which command does it, ask the CLI directly:

garmin-pp-cli which "<capability in your own words>"

which resolves a natural-language capability query to the best matching command from this CLI's curated feature index. Exit code 0 means at least one match; exit code 2 means no confident match — fall back to --help or use a narrower query. --json (and other machine formats) keep that exit-2 contract and write {"matches":[]} on stdout so agents can inspect the envelope without treating a miss as success.

Recipes

Before you run history

garmin-pp-cli history --since all --dry-run

The dry run prints the same estimate the real run prints before its expensive half, and makes no request at all, so read it before committing to the fill. What it is counting: the daily series start at 2007-01-01, the oldest day Garmin answers for whatever the account's age, and the activity feed starts at the account's first activity; one request covers a 28-day or 364-day window of a daily series, one activity for each of the three per-activity fetches, or one signal day for each of the four per-day series, and each request takes about 0.4 s. An account worn daily for years comes to several thousand requests, about an hour, and a few hundred MB on disk; a sparse account is about a thousand requests and under ten minutes. The per-activity fetches — detail, splits and heart-rate zones — cover only the activities that start on or after the day the run starts from, so a later, deeper --since extends them downward along with the daily series. When the estimate reads acceptable, start the fill with garmin-pp-cli history --since all.

A quarter of sleep, narrowed to the numbers that answer the question

garmin-pp-cli insights sleep --days 90 --agent --select period,coverage.nights_with_stats,duration_minutes.avg,score.avg,trend_vs_prior_period.duration_minutes.delta_pct

Ninety nights come out of the archive in one local read; --select keeps five numbers instead of the full nightly payload, and coverage.nights_with_stats says how many nights actually fed the averages.

Four weeks of training load and the zone split behind it

garmin-pp-cli insights training --days 28 --agent --select totals,by_type.type,by_type.duration_minutes,hr_zones

Totals, a per-type breakdown and the heart-rate-zone split in one read; drop the --select to also see coverage.activities_with_zones, which is how many of those activities were recorded with a strap.

Join two series on a value read out of the JSON

garmin-pp-cli sql "WITH days AS MATERIALIZED (SELECT substr(json_extract(data,'$.startTimeLocal'),1,10) AS day, count(*) AS activities, round(sum(json_extract(data,'$.duration'))/3600.0,1) AS hours FROM resources WHERE resource_type='activities' GROUP BY 1), steps AS MATERIALIZED (SELECT id AS day, json_extract(data,'$.totalSteps') AS steps FROM resources WHERE resource_type='steps') SELECT d.day, d.activities, d.hours, s.steps FROM days d JOIN steps s ON s.day = d.day ORDER BY d.day DESC LIMIT 14" --agent

Every archived series lives in one table, resources, keyed by (resource_type, id), so joining two of them is a self-join. When the join keys on a value read out of the JSON — here the activity's own start day against the step series' date id — give each side its own WITH ... AS MATERIALIZED (...) CTE: without MATERIALIZED, SQLite inlines the CTE and re-parses every JSON row once per joined pair. sql accepts a single SELECT or WITH statement and runs it against a handle opened read-only, so no query can change the archive whatever it says. Run garmin-pp-cli sql "SELECT resource_type, count(*) FROM resources GROUP BY 1" first to see which series this archive holds. analytics --type activities --group-by activityName stays the shorthand when one top-level field of one series is all you need, and search <text> --type activities --data-source local is the full-text equivalent over the same rows.

Connect a second household account

garmin-pp-cli auth login --email other@example.com --home ~/.local/share/garmin-homes/other

The second account gets its own home: --home for a single invocation, GARMIN_HOME to make it durable for a session.

Auth Setup

Garmin issues no personal API key, so garmin-pp-cli auth login --email you@example.com is the way in: it signs the browser out of Garmin, opens Garmin's own sign-in page, catches the one-time ticket on a loopback port on 127.0.0.1, and then asks Garmin which account it just authenticated. If that is not the address you passed, nothing is written to disk. Your password never reaches this CLI, and the refresh token keeps the session alive afterwards without another browser visit. A token supplied in GARMIN_ACCESS_TOKEN or GARMIN_TOKEN is used as-is instead: never refreshed, and not this home's stored chain. One Garmin account per home; auth status --verify asks Garmin which one this home holds. garmin-pp-cli auth logout clears that home's stored chain.

Run garmin-pp-cli doctor to verify setup.

Agent Mode

Add --agent to any command. Expands to: --json --compact --no-input --no-color.

Global format flags share one contract on promoted, novel, sync, and --deliver paths:

• --json — one JSON document on stdout (sync progress events go to stderr)

• --compact — keep identity/status/timestamp fields; does not change the document vs stream shape

• --csv / --plain — tabular rows (collection envelopes unwrap to the row array)

• --quiet — one identity value per row, no envelope

• Pipeable — JSON on stdout, errors on stderr

• Filterable — --select keeps a subset of fields. Dotted paths descend into nested structures; arrays traverse element-wise. Critical for keeping context small on verbose APIs:

  garmin-pp-cli activities list --agent --select activityId,activityName,activityType

• Previewable — --dry-run shows the request without sending

• Offline-friendly — sync/search commands can use the local SQLite store when available

• Non-interactive — never prompts, every input is a flag

• Read-only — do not use this CLI for create, update, delete, publish, comment, upvote, invite, order, send, or other mutating requests

Response envelope

Commands that read from the local store or the API wrap output in a provenance envelope:

{
  "meta": {"source": "live" | "local", "synced_at": "...", "reason": "..."},
  "results": <data>
}

Parse .results for data and .meta.source to know whether it's live or local. A human-readable N results (live) summary is printed to stderr only when stdout is a terminal AND no machine-format flag (--json, --csv, --compact, --quiet, --plain, --select) is set — piped/agent consumers and explicit-format runs get pure JSON on stdout.

Paths and state

Agents should treat the CLI's path resolver as part of the runtime contract:

• Use --home <dir> for one invocation, or set GARMIN_HOME=<dir> to relocate all four path kinds under one root.

• Use per-kind env vars only when a specific kind must diverge: GARMIN_CONFIG_DIR, GARMIN_DATA_DIR, GARMIN_STATE_DIR, GARMIN_CACHE_DIR.

• Resolution order is per-kind env var, --home, GARMIN_HOME, XDG (XDG_CONFIG_HOME, XDG_DATA_HOME, XDG_STATE_HOME, XDG_CACHE_HOME), then platform defaults.

• config contains settings like config.toml and profiles. data contains credentials.toml, data.db, cookies, and auth sidecars. state contains persisted queries, jobs, and teach.log. cache contains regenerable HTTP/cache files.

• Stored secrets live in credentials.toml under the data dir. Existing legacy config.toml secrets are read for compatibility and leave config.toml on the first auth write.

• Run garmin-pp-cli doctor --fail-on warn to surface path and credential-location warnings. agent-context exposes a schema v4 paths block for agents that need the resolved dirs.

• For MCP, pass relocation through the MCP host config. The MCP binary does not inherit CLI flags:

  {
    "mcpServers": {
      "garmin": {
        "command": "garmin-pp-mcp",
        "env": {
          "GARMIN_HOME": "/srv/garmin"
        }
      }
    }
  }

Fleet precedence: an inherited per-kind env var overrides an explicit --home for that kind. Use GARMIN_HOME or per-kind vars as durable fleet levers, and use --home only for a single invocation. Relocation is not reversible by unsetting env vars; move files manually before clearing GARMIN_HOME, or doctor will not find credentials left under the former root.

Automatic learning

This CLI ships a self-capturing learning loop. The CLI does its own bookkeeping: every invocation is journaled locally, a failed flag followed by a corrected retry auto-derives a flag_alias candidate, and a teach on a query family without a playbook auto-synthesizes a playbook_candidate from the session's journal. Your job is judgment only: recall first, act on surfaced candidates, teach the final answer, playbook amend when you observe a correction. You never record failures by hand.