gtapps/ha-morning-brief

HA Morning Brief

A house-focused morning brief that combines live HA state with hermit session context. Designed to run as the morning-brief routine at start of day.

When both claude-code-hermit and claude-code-homeassistant-hermit are installed and morning-brief is enabled, this skill subsumes /claude-code-hermit:brief --morning — operators should disable the core morning routine to avoid duplicate notifications. For hermits without HA, /claude-code-hermit:brief --morning remains the standalone path.

Steps

1. Time & context — Call GetDateTime for current time. Read .claude-code-hermit/OPERATOR.md for priorities and language preferences.

2. Fetch overnight history: Run ${CLAUDE_PLUGIN_ROOT}/bin/ha-agent-lab ha fetch-history --window-days 1. On non-zero exit, log a single line to SHELL.md ## Monitoring (history fetch failed: <stderr first line>) and skip the Overnight: section entirely — no fallback wording in the brief. On success, read .claude-code-hermit/raw/snapshot-ha-history-1d-latest.json.

3. Live house snapshot — Call GetLiveContext. Extract and organize:

   • Presence (who is home/away)
   • Lights still on (unexpected at morning time?)
   • Cover/blind positions
   • Climate: indoor temps, HVAC mode
   • Any devices unavailable or in error state
   • Security: alarm state (read-only)

4. Overnight highlights (only when 1d history artifact is present): read .claude-code-hermit/raw/snapshot-ha-normalized-latest.json for silence_summary.silent_event_sensors, then surface 1–3 highlights for the Overnight: section using only honest signals:

   • Top-active entity between 00:00–06:00: scan entity_aggregates[*].hour_histogram[0:7] for the highest sum across all entities. If any non-trivial activity (sum > 0), emit one line — e.g., "light.kitchen — 12 state changes between 00:00 and 06:00".
   • Stuck event sensors: from silence_summary.silent_event_sensors, emit sensors still silent — e.g., "binary_sensor.motion_corridor — no events for 14 days".
   • HVAC duration: for each climate.* entity, read state_durations from the history artifact. If state_durations["heat"] or state_durations["cool"] ≥ 1800 seconds (30 min), emit one line — e.g., "climate.heat_pump — heated ~Xh overnight" where X is derived directly from state_durations. Never substitute event count for active hours.

5. Energy snapshot — From the live context, pull current power draw and any energy sensors. Compare with known baselines from memory if available. Flag anything unusual (e.g., high overnight consumption).

6. Context freshness — Check .claude-code-hermit/raw/snapshot-ha-context-latest.json modification time. If older than 24h, note it as stale.

7. Overnight activity — Read .claude-code-hermit/sessions/SHELL.md. Scan both the Monitoring section and the Findings section (last 20 lines combined). In newborn-phase hermits (< 3 days old), pattern observations land in Findings as Noticed: <pattern> entries — include those. Surface any alerts or notable patterns found overnight.

8. Cost-spike check — Read .claude-code-hermit/state/reflection-state.json if it exists. Look for any cost_spike entry with a timestamp within the last 24 hours. If found, include a "Cost alert" bullet in the brief with the flagged amount.

8a. Pending updates: Run ${CLAUDE_PLUGIN_ROOT}/bin/ha-agent-lab ha updates --digest and capture stdout. Branch on its content (the command always exits 0 — never branch on exit code):

• Contains (skipped: — log a single line to SHELL.md ## Monitoring (updates fetch failed: <detail after "skipped:">) and omit the Updates: section entirely.
• Contains (no updates pending) — omit the Updates: section entirely.
• Otherwise — render the digest lines as the Updates: section, translating tier labels into the operator's language. For each [tier] Title: installed → latest line (skip the + N more … collapse line and the [hacs] aggregate line), Glob .claude-code-hermit/proposals/PROP-*.md for a [ha-update] proposal whose title carries the same tier and target version (e.g. [ha-update] HA Core → 2026.7.1) and append (PROP-NNN); if none matches yet, render the line without an id. Step 9 reuses this glob result when this branch ran it.

9. Pending work — Scan for:
   • Glob .claude-code-hermit/proposals/PROP-*.md (reuse step 8a's glob result if its Otherwise branch already ran it; otherwise run the Glob here) — read status from each, list any pending proposals. Exclude a pending proposal whose title starts with [ha-update] only when the Updates: section is present (it surfaces there instead); when step 8a omitted the Updates: section (skipped fetch or no updates pending), keep [ha-update] proposals in this list so they are not lost
   • Check if .claude-code-hermit/sessions/NEXT-TASK.md exists (queued task)
   • Read .claude-code-hermit/cost-summary.md if it exists — include yesterday's cost

<!-- keep in sync with plugins/claude-code-hermit/skills/brief/SKILL.md — same MP lifecycle protocol -->

9a. Micro-proposals lifecycle — age the queue in one call: .claude-code-hermit/bin/hermit-run proposal micro .claude-code-hermit brief-cycle. This runs core's writer through the project-resident bin/hermit-run, which resolves core's plugin root (a static ../claude-code-hermit/… path can't — HA's ${CLAUDE_PLUGIN_ROOT} is <cache>/<marketplace>/claude-code-homeassistant-hermit/<version>/ and the version segment isn't knowable from skill text). It performs the whole follow_up_count 0/1/2+ lifecycle atomically (re-nudges count-1 entries, expires count-2+ entries, records each expiry) and prints one JSON line {"new":[…],"renudged":[…],"expired":[…]}. Never hand-edit state/micro-proposals.json. Render from that verdict:

• Each new entry: include in Awaiting decision: output (see Output Format).
• Each renudged entry: include in Awaiting decision: with softer framing: "Still waiting on MP-YYYYMMDD-N: [question] (ignore again to drop it)".
• expired entries were dropped this cycle — do not surface them.
• If new and renudged are both empty: skip this step entirely.

10. Compose brief — Write a concise morning brief in the operator's language (from OPERATOR.md preferences). Use the format below.

11. Write to compiled/ — Write the composed brief to .claude-code-hermit/compiled/brief-morning-<YYYY-MM-DD>.md with frontmatter:

title: "Morning Brief — <YYYY-MM-DD>"
type: brief
created: <ISO8601>
session: <session_id from runtime.json, or null if absent>
tags: [morning-brief, ha]

Then append the following line to .claude-code-hermit/sessions/SHELL.md under a ### Artifacts produced this session subsection in ## Monitoring (create the subsection if absent):

- [[compiled/brief-morning-<YYYY-MM-DD>]]

This citation is lifted into ## Artifacts when /claude-code-hermit:session-close archives the session.

Output Format

Good morning! Home - [date]

Current state:
- [presence, lights, climate, covers - concise bullets]

Overnight:
- [top-active entity, stuck sensors, HVAC duration — omit section when history unavailable]

Energy:
- [current draw, notable consumption]

Alerts:
- [devices offline, unusual states, or "All clear"]

Updates:
- [tier] Title: installed → latest (PROP-NNN if a matching proposal exists)
- [N HACS updates pending]
- [Omit section entirely when none pending or the fetch was skipped.]

Pending:
- [proposals, queued tasks, or "Nothing pending"]

Awaiting decision:
- [`new` entry: "MP-YYYYMMDD-N (tier N): [question] — Reply `MP-YYYYMMDD-N yes` or `MP-YYYYMMDD-N no`"]
- [`renudged` entry: "Still waiting on MP-YYYYMMDD-N: [question] (ignore again to drop it)"]
- [Omit section entirely when the step-9a verdict's `new` and `renudged` are both empty.]

Cost:
- [yesterday's cost if available; cost-spike alert if flagged by reflect]

Today's priorities:
- [from OPERATOR.md Current Priority, filtered to actionable items]

Adapt the greeting and section headers to the operator's configured language. Keep the entire brief under 25 lines — strip lower-priority lines (e.g., Cost when no spike) if the Overnight section pushes over the cap. Awaiting decision: lines are final and non-droppable — strip from Energy, Cost, Today's priorities, and Updates (in that order) before touching MP lines.

Delivery

• If invoked as a routine, or config.always_on is true in .claude-code-hermit/config.json: deliver the composed brief via the Operator Notification protocol in CLAUDE.md (core resolves the channel and falls back to push / SHELL.md logging when no channel is reachable). The terminal is unmonitored in always-on mode — never gate delivery on session_state. For the push-fallback branch, condense to a single line (per § Operator Notification push format): lead with any overnight anomaly or open Awaiting decision: count, then energy/cost if flagged. Example: House OK overnight, 2 awaiting decision, 14kWh — open CC to view.
• Otherwise (invoked on demand in an interactive session): output to terminal.
• Never include secrets, tokens, or internal file paths in the brief.