claude-code-hermit

Reflect on recent work and propose improvements if patterns are noticed.

Schedules routines via one persistent Monitor subprocess (zero-token skips); CronCreate fallback where Monitor is unavailable. heartbeat-restart stays a CronCreate re-arm anchor.

Initializes the autonomous agent in the current project. Creates the state directory, templates, OPERATOR.md, and config.json. Appends session discipline to CLAUDE.md. Detects installed hermits. Run once per project, like git init.

Morning house brief — live status, overnight anomalies, energy snapshot, pending proposals, and today's priorities. Runs as a daily routine or on demand.

Evening house brief — end-of-day security check, device status, and energy snapshot. Runs as a daily routine at 22:30 or on demand.

Read site deployment logs, server logs, site application/nginx logs, background process logs, and specific deployment logs from Laravel Forge. Includes a triage mode that bundles recent logs with deployment history for rapid incident analysis. Triggers on "show logs", "read logs", "deployment log", "server log", "what happened to the deployment", "triage failing site".

Bump a plugin's version, write a detailed changelog entry for the upgrade skill to consume, and commit+push. Takes a plugin slug argument identifying which plugin under `plugins/` to release. Use this skill whenever the user says "release", "version bump", "cut a release", "changelog and push", or finishes a set of changes and wants to ship them. Also trigger when the user says "do the release thing" or asks to prepare changes for hermits to pick up.

Opt-in advanced Docker security hardening beyond the v1.0.26 baseline — LAN containment with DNS policy, resource bounds with kernel-hygiene sysctls, and a boot-time plugin install audit log. Applied as a reversible docker-compose overlay (does not touch the base compose file) and verified against the live container. Run after /docker-setup; requires bridge networking.

Generates Docker scaffolding and walks the operator through the full deployment — token setup, build, start, MCP plugin configuration, workspace trust, and verification. Offers to back up and overwrite existing Docker files. Run after /hatch.

Returns a twenty-four-check health report on the hermit installation (runtime, config, hooks, state-file integrity, cost, proposals, deps, version currency, permissions, docker, archival, reflect loop, scheduler, watchdog, context age, opus-wake spend, routine cost, heartbeat, routine monitor, raw storage size, sibling-plugin credential expiry, model pricing, channel liveness, context scan). Use when diagnosing an install, before a release, or after suspicious behavior. Activates on messages like "/hermit-doctor", "health check", "diagnose the hermit", "what's wrong", "run diagnostic".

Executes the heartbeat checklist from HEARTBEAT.md. Reads the checklist, evaluates each item, and reports findings or acknowledges with HEARTBEAT_OK. Supports run/start/stop/status/edit subcommands.

View or change hermit configuration for this project. Manages model, channels, morning brief, heartbeat, routines, idle behavior, compaction thresholds, Docker packages, and unattended mode.

Returns a 5-line executive summary of recent work. Checks active session first, falls back to latest report. Activates on messages like "brief", "what happened", "morning update", "overnight summary", "status", "progress", "what are you working on", "how's it going".

Generate the weekly review report for the current ISO week. Writes to .claude-code-hermit/compiled/review-weekly-YYYY-Www.md (dev-facing detail) and sends a plain-language channel summary (Delivered / Decisions / Waiting on you / Spend). Runs every Sunday at 23:00 via routine.

Initializes or resumes a work session. Loads context from OPERATOR.md and SHELL.md, orients the agent, and establishes what to work on. Use at the beginning of every work session.

Evolves hermit configuration and templates after a plugin update. Detects version gaps and runs the upgrade (migrations, templates, new settings) in an isolated subagent. Run after updating the plugin.

Apply a Home Assistant update surfaced by ha-update-check, when the operator accepts a [ha-update] proposal. Enforces the tier rule -- add-ons/HACS may auto-apply, Core/OS/Supervisor always wait for an explicit operator go-ahead.

Detect pending Home Assistant updates (Core, OS, Supervisor, add-ons, HACS) from the update.* domain and surface them as actionable proposals. Runs daily as a scheduled check via reflect --scheduled-checks.

One-time Home Assistant setup for this hermit. Configures HA access, connects to the official Home Assistant MCP Server integration, and verifies both the CLI and HA MCP. Run once per project after /claude-code-hermit:hatch.

Handles inbound messages from Claude Code Channels (Telegram, Discord, webhooks) with session context awareness.

Accept, defer, dismiss, or resolve a proposal. For accepted proposals, asks how to proceed: start implementing now, create a session task, or note for manual implementation. Activates on messages like "accept PROP-", "dismiss PROP-", "defer PROP-", "resolve PROP-".

Creates a proposal for a high-leverage improvement discovered during work. Only for ideas with real impact — not trivial fixes. Use when you discover something worth operationalizing.

Pre-release gate for the monorepo — establishes per-plugin release boundaries (each plugin's last tag → HEAD), audits every changed plugin's CHANGELOG against the actual code diff, then hands off to the native `/code-review high --fix` for the deep correctness pass. Use this whenever the operator is about to ship and says "pre-release review", "review before I release", "review what's shipping", "audit the release", "check the release window", "is this ready to tag", or asks to sanity-check accumulated unreleased work before `/release` or `/fleet-release`. Runs BEFORE the release, never mutates version/tags. Trigger even when phrased loosely, as long as the intent is a last look before shipping.

Start or resume a work session with full context loading and work tracking. Use at the beginning of work.

The 'why is my bill high' skill: audits where spend goes across token categories (cache_read, cache_write, output, input) and per-session attribution. Invoke for questions like 'cost breakdown', 'which sessions are expensive', 'what's driving spend', or 'cold start costs'. Not for week-over-week trend lines (use hermit-evolution) or simple total-spend summaries.

Lists all proposals with their status, source, and age. Shows auto-detected proposals prominently. Activates on messages like "what have you noticed", "any improvements", "any proposals", "show proposals".

Guided channel activation for local/tmux users — installs the plugin, configures the bot token in the project-local state dir, and walks through pairing. Run after hatch or hermit-settings to activate a configured channel.

Closes the current work session with a structured handoff. Archives the session report and prepares for the next session. Activates on messages like "I'm done", "wrap it up", "that's it for now", "done for today", "close the session". Also runs the midnight `--scheduled` decision (close now, queue, or noop) fired by the `daily-auto-close` routine.

On-demand hermit-voice brainstorm — synthesizes memory, capabilities, recent artifacts, and codebase shape into at most 2 capability ideas, each gated by proposal-triage before becoming a PROP. Invoke when the operator explicitly asks to brainstorm capabilities or ideate, e.g. "brainstorm capabilities", "what could you be doing for me?", "any capability ideas?". Never runs autonomously.

Start-of-work triage in this monorepo. Trigger on: a GitHub issue number ("tackle issue 99", "work on #42", bare #N), pasted proposal/spec text ("investigate this", "pressure-test this", "scope this"), or picker phrasing ("tackle the next issue", "what should I work on"). Read-only investigation that tries to falsify the premise, then recommends SHIP / SHIP WITH CAVEAT / DEFER / SKIP. On SHIP verdicts for a real issue number it offers branch checkout + task seeding; --investigate-only suppresses that handoff. Not for PR review or work already in progress.

One-time fitness hermit setup. Configures Strava MCP access, drops routine prompt templates, and wires routines into config.json. Run once per project after /claude-code-hermit:hatch.

One-time setup for hermit-scribe — appends the Issue Filing block to CLAUDE.md/CLAUDE.local.md and seeds an auto-mode environment entry naming api.github.com as a trusted, always operator-confirmed destination. Run once per project; re-run to refresh after an upgrade.

Pre-release docs-drift audit for the monorepo — reads every plugin's [Unreleased] changelog section and checks whether the documentation surfaces (plugin README, docs/, plugin CLAUDE.md, root README, root CLAUDE.md) still tell the truth after those changes. Reports meaningful drift with concrete proposed edits and asks before applying anything. Use this whenever the operator is preparing a release and says "docs drift", "check the docs", "are the docs up to date", "docs audit", "did the docs keep up", "check documentation before release", or wants to know whether READMEs/docs match what's about to ship. Complements /pre-release-review (changelog vs code); this skill covers changelog vs docs. Trigger even when phrased loosely, as long as the intent is verifying docs against accumulated unreleased changes.

Show alert state, proposal queue depth, routine engagement, knowledge state, channel availability, plus fragile zones, stale accepted proposals, and recent learnings. Activates on messages like 'health check', 'how's the hermit', 'is anything broken', 'hermit health', 'system health', 'anything wrong', 'hermit status', 'hermit brain', 'are routines running', 'what's stuck', 'any fragile zones', 'show me what's blocked', 'recent learnings', 'what have you learned lately', 'where are the weak spots', 'check knowledge', 'lint knowledge', 'knowledge health'.

Per-activity coaching analysis from Strava. Detects interval vs steady-state sessions and road vs trail terrain, branching the metrics accordingly. Computes zone breakdown and recovery estimate; saves a compiled artifact and returns a compact summary. Run after a workout or to retro-analyze a specific activity.

Captures RPE (1-10) and subjective notes when the operator replies to a strava-sync notification on any configured channel. Triggers on inbound channel messages matching RPE grammar while state/strava-pending-rpe.json has a sync from the last 24h. Channel-agnostic.

Manually record RPE and subjective notes for a specific Strava activity. Use for backfilling, correcting entries, or rating a non-latest activity that the auto-capture skill skipped.

Interval scheduled check — detects multi-session cardiac-drift trends across recent compiled activity notes. Runs weekly via the scheduled-checks routine; findings are routed through the proposal pipeline as Evidence Source: scheduled-check/weekly-coaching-patterns.

Activate the dev hermit in the current project. Appends the dev safety rules to CLAUDE.md, installs the git-push-guard hook at strict profile by default, and offers companion plugins. Run once per project after /claude-code-hermit:hatch; re-run to update individual settings.

Operator-sanctioned alternative to bare `git push` from agent context. Open a PR from the current feature branch with a body assembled inline from commit history, the test command's last output, screenshots, and an optional project PR template. Refuses on protected branches, dirty trees, or zero commits ahead. Run as the final step of a ticket.