gtapps/hatch

Hatch — claude-code-fitness-hermit

Idempotent setup wizard for the fitness plugin. Run after /claude-code-hermit:hatch has already been completed.

---

Step 1 — Prerequisite check

Check whether .claude-code-hermit/config.json exists.

If it does not:

"The base hermit is not set up in this project yet. Run /claude-code-hermit:hatch first, then return here."

Use AskUserQuestion: "Would you like to run /claude-code-hermit:hatch now? (yes / no)"

• yes → Follow the domain hatch continuation protocol (documented in claude-code-hermit:hatch):
  1. Write .claude-code-hermit/state/hatch-resume.json with { "skill": "claude-code-fitness-hermit:hatch" }.
  2. Print: "(If setup doesn't continue automatically when core finishes, re-run /claude-code-fitness-hermit:hatch.)"
  3. Invoke /claude-code-hermit:hatch via the Skill tool — terminal action, stop after the call.
• no → stop.

If it does exist, run .claude-code-hermit/bin/hermit-run domain-hatch preflight claude-code-fitness-hermit and parse the JSON verdict. Branch on action:

• upgrade-core-package / upgrade-core-applied → relay the remedy string verbatim to the operator and stop.

• verify → say:

  "claude-code-fitness-hermit {self_version} is already installed. Skip to Step 7 to re-verify the installation, or reply 'full' to re-run the full wizard."

  Use AskUserQuestion: "(verify / full)" — verify → skip to Step 7; full → continue from Step 2.

• full → continue from Step 2.

• ok: false → relay message and stop.

---

Step 2 — .env verification

IMPORTANT: Do NOT use grep, cat, echo, or any Bash command to read .env. Three of the four required variables contain the literal string TOKEN in their name, which triggers the base hermit's deny-patterns hook on any Bash command argument. Use the Read tool only.

Tell the operator:

"This plugin needs four Strava OAuth credentials in .env. If you haven't done this yet:

1. Create .env at the project root
2. Open .env and fill in all four values from your Strava developer app:
   • STRAVA_CLIENT_ID — numeric app ID
   • STRAVA_CLIENT_SECRET — app secret
   • STRAVA_ACCESS_TOKEN — initial access token
   • STRAVA_REFRESH_TOKEN — refresh token

See https://developers.strava.com/docs/authentication/ to create a Strava app if you haven't. Reply 'done' when the file is filled in, or 'abort' to stop."

Use AskUserQuestion: "(done / abort)"

• abort → stop.
• done → continue.

Use the Read tool to read .env. Parse each line of the form KEY=VALUE. Verify that all four of these keys are present and their values are not empty and not replace_me:

• STRAVA_CLIENT_ID
• STRAVA_CLIENT_SECRET
• STRAVA_ACCESS_TOKEN
• STRAVA_REFRESH_TOKEN

If any value is missing or still set to replace_me, report which ones and loop back to the AskUserQuestion above. Do not proceed until all four are valid.

---

Step 3 — MCP registration

Step 3.0 — Detect an ambient Strava MCP server. The operator may already run a Strava MCP server (user-scoped, or configured by another tool). Run claude mcp list (Bash) and scan for a server that talks to Strava — a name containing strava, or an entry whose command/args reference a Strava MCP package (e.g. strava-mcp-server).

• None found → proceed to install the bundled server below (the default path).
• One found → ask with AskUserQuestion (header: "Strava MCP"): Reuse the existing <name> server (recommended — no duplicate) / Install the bundled @r-huijts/strava-mcp-server.
  • Reuse → skip the .mcp.json write entirely. Record which server key the skills should target (if it is not strava, note in the final report that skill/settings matchers assume the key strava, so the operator should either rename their server to strava or accept that the fitness skills call mcp__strava__*). Continue to Step 4.
  • Install bundled → proceed below.

This is a local reuse-vs-install choice only — do not attempt to reconcile or edit the operator's other MCP configs.

Using the four values you parsed from .env in Step 2 (held in working context — do not re-read .env), write the Strava MCP server entry into the project's .mcp.json.

Do not embed ${VAR} placeholders — substitute literal values. Claude Code passes MCP env blocks as literal environment variables to the child process; it does not expand shell variable syntax.

Read the project root .mcp.json (use the Read tool; treat as { "mcpServers": {} } if the file does not exist).

Check if a strava key already exists under mcpServers:

• Absent → add the entry.
• Present with ${...} placeholders → rewrite just the strava entry with literal values.
• Present with literal values → skip (already configured).

The strava entry shape:

{
  "strava": {
    "command": "npx",
    "args": ["-y", "@r-huijts/strava-mcp-server"],
    "env": {
      "STRAVA_CLIENT_ID": "<literal STRAVA_CLIENT_ID from .env>",
      "STRAVA_CLIENT_SECRET": "<literal STRAVA_CLIENT_SECRET from .env>",
      "STRAVA_ACCESS_TOKEN": "<literal STRAVA_ACCESS_TOKEN from .env>",
      "STRAVA_REFRESH_TOKEN": "<literal STRAVA_REFRESH_TOKEN from .env>"
    }
  }
}

Write the updated .mcp.json using the Write tool.

Add .mcp.json and .env to .gitignore if not already present. Read the project .gitignore, check for each entry; append any that are missing on new lines using Edit.

---

Step 4 — Drop routine prompt files

Copy the four routine prompt templates from the plugin's state-templates/compiled/ into the consumer's .claude-code-hermit/compiled/.

For each of the four files:

• ${CLAUDE_PLUGIN_ROOT}/state-templates/compiled/routine-fitness-brief-morning.md
• ${CLAUDE_PLUGIN_ROOT}/state-templates/compiled/routine-fitness-brief-evening.md
• ${CLAUDE_PLUGIN_ROOT}/state-templates/compiled/routine-weekly-load-review.md
• ${CLAUDE_PLUGIN_ROOT}/state-templates/compiled/routine-monday-planning.md

Read the source file (using Read tool), then check if the destination exists (.claude-code-hermit/compiled/<filename>):

• Does not exist → write it using Write tool. Report: ✓ dropped <filename>.
• Already exists → skip (do not overwrite operator edits). Report: ⊘ skipped <filename> (already present).

---

Step 5 — CLAUDE.md / CLAUDE.local.md inject

Resolve target file: Step 1's preflight already returned target, target_file, target_default and needs_target_question.

If needs_target_question is true, ask with AskUserQuestion (header: "Visibility") — target_default at position 0 with (recommended): .local files (gitignored — operator-personal) / Committed files (shared with teammates). Then record it:

.claude-code-hermit/bin/hermit-run domain-hatch ensure-target claude-code-fitness-hermit --target <choice>

Then write the block:

.claude-code-hermit/bin/hermit-run domain-hatch sync-block claude-code-fitness-hermit

It appends the <!-- claude-code-fitness-hermit: Fitness Workflow --> block when the marker is absent (creating target_file if needed) and skips when it is already present; hermit-evolve handles block replacement on upgrade.

Stray-block migration (block stranded in the non-target file after a target flip) is handled one-shot by the Upgrade Instructions in this version's CHANGELOG entry, executed by hermit-evolve Step 7. Hatch itself stays focused on target-aware setup.

---

Step 6 — Knowledge-schema extension

Read .claude-code-hermit/knowledge-schema.md.

Check if activity-streams: is present in the file. This string only appears in the Raw Captures section, which is the last thing written — so its presence means both blocks were fully written on a prior run.

If absent, append the following block under ## Work Products (create the section header if the base schema only has a template stub):

- weekly-plan: weekly training structure suggestion (7-day breakdown). Triggered by monday-planning routine (Mon 09:30). location: compiled/weekly-plan-<YYYY-MM-DD>.md
- weekly-summary: week-over-week training load review. Triggered by weekly-load-review routine (Sun 18:00). location: compiled/weekly-summary-<YYYY-MM-DD>.md
- recovery-assessment: recovery indicators from recent activity data. Triggered by operator request or an evening-brief flag. location: compiled/recovery-assessment-<YYYY-MM-DD>.md
- fitness-snapshot: current fitness state snapshot. Triggered by operator request. location: compiled/fitness-snapshot-<YYYY-MM-DD>.md
- activity-note: per-activity coaching analysis. Triggered by activity-deep-dive skill. location: compiled/activity-<id>-<YYYY-MM-DD>.md

And under ## Raw Captures (create if absent):

- activity-fetch: raw activity list from Strava. Feeds weekly-plan, weekly-summary, recovery-assessment. Retention: 3 days. location: raw/activity-fetch-<date>.json
- activity-streams: HR/pace/power time-series for a specific activity. Feeds recovery-assessment, activity-note, fitness-snapshot. Retention: 7 days. location: raw/activity-streams-<id>-<date>.json

If already present: skip (idempotent).

Use Edit to make the changes.

---

Step 7 — Stamp and register in config.json

Re-read .claude-code-hermit/config.json now — the wizard has been running since Step 1 and the on-disk file may have changed. Apply the merges below to that fresh copy.

7a — Stamp version

Set _hermit_versions["claude-code-fitness-hermit"] to self_version from Step 1's preflight.

If the key already exists: update the value. If absent: add it alongside the existing _hermit_versions["claude-code-hermit"] entry.

7b — Merge routines

In the routines array, check for each of these four IDs. For any that are absent, add the entry. For any that are present (by id), skip (do not clobber existing operator edits).

{
  "id": "morning-brief",
  "schedule": "30 7 * * *",
  "skill": "claude-code-hermit:session-start",
  "enabled": true,
  "run_during_waiting": true,
  "prompt_file": "compiled/routine-fitness-brief-morning.md"
},
{
  "id": "evening-brief",
  "schedule": "30 21 * * *",
  "skill": "claude-code-hermit:session-start",
  "enabled": true,
  "run_during_waiting": true,
  "prompt_file": "compiled/routine-fitness-brief-evening.md"
},
{
  "id": "weekly-load-review",
  "schedule": "0 18 * * 0",
  "skill": "claude-code-hermit:session-start",
  "enabled": true,
  "run_during_waiting": true,
  "prompt_file": "compiled/routine-weekly-load-review.md"
},
{
  "id": "monday-planning",
  "schedule": "30 9 * * 1",
  "skill": "claude-code-hermit:session-start",
  "enabled": true,
  "run_during_waiting": true,
  "prompt_file": "compiled/routine-monday-planning.md"
}

7c — Merge scheduled_checks

In config.scheduled_checks, check for an entry with id: "weekly-coaching-patterns". If absent, append it. If present (by id), skip — do not clobber existing operator edits.

{"id": "weekly-coaching-patterns", "plugin": "claude-code-fitness-hermit", "skill": "claude-code-fitness-hermit:weekly-coaching-patterns", "enabled": true, "trigger": "interval", "interval_days": 7}

No prompt needed — this is a read-only analysis. The core daily scheduled-checks routine picks it up; interval_days: 7 gates cadence. Findings surface as proposals automatically via the existing pipeline.

Write the updated config.json using Write tool (full file replacement to ensure valid JSON).

7d — Auto-mode environment seed

Run bun ${CLAUDE_PLUGIN_ROOT}/scripts/automode-env.ts .claude/settings.local.json — always .claude/settings.local.json, regardless of hatch_target: Claude Code's auto-mode classifier reads autoMode config only from local/user scope, never a committed project .claude/settings.json. This names www.strava.com as a trusted external service, so the classifier stops treating the nightly evening-brief routine's read-only fetches as unrecognized outbound calls. Additive and idempotent; safe to re-run on every hatch. No prompt needed.

---

Step 8 — Final report

Print a structured summary:

claude-code-fitness-hermit {version} setup complete.

Installation summary:
  ✓ Prerequisite: claude-code-hermit {base_version} confirmed
  ✓ .env: all four Strava credentials present
  ✓ .mcp.json: strava server entry written (or was already present)
  ✓ .gitignore: .mcp.json and .env covered
  ✓ Routine prompts: {N}/4 dropped, {M}/4 already present
  ✓ CLAUDE.md: Fitness Workflow block injected (or was already present)
  ✓ knowledge-schema.md: fitness types added (or were already present)
  ✓ config.json: _hermit_versions stamped, {K}/4 routines added, weekly-coaching-patterns check registered

Manual steps remaining:
  - Restart Claude Code so the `strava` MCP server loads from .mcp.json
  - Approve the `strava` server when prompted on first use
  - Run /mcp to confirm `strava` is connected
  - Verify connectivity: call mcp__strava__check-strava-connection

Go always-on (recommended):
  - Docker:     /claude-code-hermit:docker-setup
      Builds the container and walks you through channel pairing in one go.
  - Bare tmux:  .claude-code-hermit/bin/hermit-start
      For channels (Discord/Telegram) with tmux, run
      /claude-code-hermit:channel-setup first.

Prefer to test interactively first? After restarting, run:
  /claude-code-hermit:hermit-routines load
    — activates the four fitness routines in the current Claude session.

The always-on runtime activates routines automatically — the interactive
steps are only for a test drive before handing over to the runtime.

Installed skills:
  /claude-code-fitness-hermit:fitness-brief             — daily morning/evening brief (--morning|--evening|--slot)
  /claude-code-fitness-hermit:activity-deep-dive        — per-activity coaching analysis
  /claude-code-fitness-hermit:weekly-coaching-patterns  — weekly cardiac-drift trend check (scheduled, interval_days: 7)

Installed subagent:
  @claude-code-fitness-hermit:strava-data-cruncher — bulk Strava data aggregation (Haiku)

Security reminder: .env and .mcp.json contain real Strava credentials. Both are gitignored — verify before any git push.

---

Docker network requirements

Read by /claude-code-hermit:docker-security when the operator enables LAN containment + DNS policy. Each entry is surfaced as a per-entry confirmation prompt; nothing here is auto-applied.

Domains (DNS allowlist)

• strava.com

server=/strava.com/... matches both www.strava.com (web) and api.strava.com (API) via dnsmasq's subdomain-match semantics. The Strava MCP server (@r-huijts/strava-mcp-server via npx) also needs npmjs.org, but that's already in the wizard's seed allowlist for npm/plugin installs.

LAN allowlist suggestions

(none — Strava is a cloud service)