larsboes

@larsboes source repo

29 skills
0 followers
6 hours ago last updated

▌
Axon · larsboes bundle
Operates Axon and guides work on the Axon repository through dynamic discovery, bounded self-context, capability APIs, architecture placement, focused branches, and proportionate verification. Use when operating Axon services or feeds, diagnosing Axon, planning or coding inside the Axon repository, deciding where Axon code belongs, or handling Axon branches and PRs. Do not use for unrelated repositories, direct Home Assistant or device control, or generic prose and skill authoring with no Axon decision.

0
installs
▌
Trim · larsboes bundle
Shrinks the files that load into every session — the user and project CLAUDE.md, their @-imports, the memory index — by removing what is provably dead first and then merging, tightening or relocating the rest behind a human gate, without ever dropping a distinct directive. Use when an always-on context file has grown too big, when asked to trim, shrink or prune CLAUDE.md, AGENTS.md or a memory index, when a rule set has turned into a document, or when the session's instruction budget needs to come down. Do not use for editing ordinary documentation, for refactoring code, or for removing AI writing tells from prose.

0
installs
▌
Cv Builder · larsboes
Builds a job-tailored CV PDF from a single profile-tagged master CV file, via the local `cv` CLI — pick the closest profile tag (or ask), pick a language, run the build, hand back the PDF path. Use when the user wants a CV/resume for a specific job application, role type, or language. Do not use for writing or editing CV content itself (that's the master file, edited by hand) or for any content-generation/AI-tailoring flow — this skill only filters and renders what's already in the master file.

0
installs
▌
Travel · larsboes bundle
Drives Axon's travel capabilities as one workflow — feasible days from calendar, journey search and split tickets from transit, measured delay history from punctuality, event discovery from scouting, and plan state in trips. Use when planning a trip, searching connections or fares, checking how late a train usually runs, or turning a discovered event into an itinerary. Do not use for home network or Home Assistant work, and never work from a remembered route or port — ask the service.

0
installs
▌
Unslop · larsboes bundle
Strips the tells that make source code and web UI read as AI-generated and forces a deliberate, project-specific choice instead of the model's default average. It writes neither the code nor the design, and has no house style. Code: chat artifacts, placeholder comments, emoji, swallowed errors, generic names like process_data, plus the structural tells a linter passes (tutorial boilerplate, hallucinated APIs, over-engineering, ignoring the surrounding repo). UI: default shadcn/Tailwind, AI-purple gradients, emoji-as-icons, the hero-plus-three-cards skeleton, and the newer cream-serif-sage 'tasteful default'. Use whenever writing, generating, reviewing, refactoring, or auditing code or a web interface, especially when it 'looks AI-generated', 'reads like a tutorial', is 'too generic' or 'vibe-coded'. Trigger even if the request never says AI tell. Do not use for prose a human reads (use human-writing), thesis text (use academic-writing), or agent skills (use writing-skills).

0
installs
▌
Teach · larsboes
Optimal one-to-one teaching that grows the Knowledge vault. Use when the user wants to learn, understand, or be taught a topic ("teach me X", "I want to learn X", "/skill:teach X"). Probes the exact edge of their understanding with graded quizzes, plans a dependency path, then teaches one reasoning step at a time — writing everything durable into vault Knowledge notes through the obsidian skill, linking what they already know instead of re-teaching it. Do not use for plain question-answering, for ingesting a source into the wiki, or for vault operations with no teaching in them (obsidian).

0
installs
▌
Obsidian · larsboes bundle
Operates the user's Obsidian vault as the single durable knowledge surface — resolves the vault, gates every write on the vault contract, creates and retypes notes, runs bulk property/link sweeps, maintains the wiki layer, and recalls what earlier sessions said. Use for any work inside the vault (notes, frontmatter, MOCs, bases, canvases, clippings, session and planning notes), for "what do we know about X", and for "what did we do last week / where did we leave off". Do not use for generic markdown outside the vault, for prose style (human-writing), or for authoring skills (skill-creator).

0
installs
▌
Council · larsboes bundle
Convenes four to five subagents that hold different positions on one decision, runs them over rounds, and reports a transcript, a recommendation and the minority position. Three rounds for a decision that is expensive to reverse, one round for a sanity check. Carries five domain presets — architecture, investment, travel, security, product — and a rule for composing a council when none of them fits. Use when the request asks for a council, a debate, several perspectives, a weighing of options, or the case for and against a choice. Do not use for a pure attack on one proposal (use red-team), and do not use for a question that has one factual answer.

0
installs
▌
Harness Sync · larsboes
Answers where every Axon Pack skill is deployed across the agent harnesses installed on this machine, what has drifted from its source, and which of the two reverse moves keeps an edit that was made inside a harness. Drives tools/harnesses (list, status, drift, sync, promote, accept). Use whenever the question involves deployed skills or agents — "is everything deployed", "sync my skills", "what drifted", "which harnesses have this skill", "bring this skill into Axon", "I edited a skill in place", "why is this skill missing in pi", "deploy the pack" — and whenever a Pack, a skill or an agent file has just been added, renamed or edited in the Axon repository. Do not use for writing the content of a skill, for choosing which skill to build next, or for shrinking always-on context files.

0
installs
▌
Red Team · larsboes bundle
Attacks one proposal. Decomposes it into atomic claims, states the steelman its author would sign, then argues the strongest surviving counter-argument, with findings ranked by severity and a test that would settle each one. Use when the request asks to red-team, stress-test, poke holes in, break, challenge, play devil's advocate against, or find the weaknesses of a plan, an argument, a design or a decision that has already been made. Do not use for weighing two or more options against each other (use council), and do not use for scanning software, a network or a host for vulnerabilities.

0
installs
▌
Fritz · larsboes bundle
Queries an AVM Fritz!Box (7590/FRITZ!OS) read-only over two transports — TR-064 via fritzconnection for the authoritative host table, and the AVM JSON API (login_sid.lua + data.lua) for what TR-064 cannot see: active channel and width per band, DFS exposure, mesh topology with per-AP client counts, the timestamped WLAN event log, and subsystem load. Credentials come from a Bitwarden item at runtime, never hardcoded. Use when listing network devices, checking what is on the network, Fritz!Box device discovery, WLAN clients, guest-WiFi status, spotting an unknown device, or diagnosing a slow or unstable WLAN (channel width, DFS/radar eviction, repeater backhaul, airtime saturation). Do not use for Home Assistant entity control (use ha-cli), DNS or ad-blocking (use pihole), local subnet scanning and drift (use netmon), or non-Fritz routers.

0
installs
▌
Human Writing · larsboes bundle
Drafts, edits, reviews, and audits prose so it reads as one deliberate human voice instead of the model's default register. Combines a linter for the mechanical tells (em dash, "it's not just X, it's Y" cadence, assistant boilerplate, diction memes, bold-bullet listicles), a structural pass for the tells no regex catches (uniform rhythm, sycophancy, saying nothing at length), and a guard against the over-corrected "trying not to sound like AI" register that swaps one default for another. Facts, numbers, and citations stay invariant. Use when writing, drafting, editing, rewriting, reviewing, or auditing prose for a reader (post, email, essay, README, marketing copy), when text "sounds like ChatGPT", "reads like AI", or "is too polished", and for "de-slop this" or "make it sound like me". Trigger even if the request never says "AI tell". Do not use for authoring agent skills (use writing-skills), thesis or paper prose (use academic-writing), source code or web UI (use unslop).

0
installs
▌
Suggest Skills · larsboes bundle
Reads this machine's own prompt history and installed skills, finds the work that keeps recurring and the places the user kept having to correct the agent, and returns a ranked shortlist of skills worth building — each with the evidence that produced it. Read-only and proposal-only — it never writes, edits or installs a skill. Use when the request is "what skills should I build", "am I missing a skill", "skill gap", "based on my recent work, what would help", or when a review of a skill library asks what is absent rather than what is wrong. Do not use for writing or auditing one skill (that is the skill-authoring skill), and do not use it to decide whether an existing skill is any good.

0
installs
▌
Ha CLI · larsboes bundle
Queries and controls an explicitly configured Home Assistant instance and can write a private inventory into the active overlay. Use for Home Assistant state, configuration, integration, service-call, reload, or inventory operations. Never place returned instance data in public Axon.

0
installs
▌
Netmon · larsboes bundle
Monitors a home LAN and its traffic from the local host — active host discovery with MAC and vendor, per-host open-port scans, a saved baseline plus drift diff (new/missing devices), WAN throughput totals, and DNS-egress visibility (which domains each device phones home to) via the resolver's query log. Subnet is auto-derived from the primary interface or set by env/flag; Fritz!Box throughput credentials come from a Bitwarden item at runtime, never hardcoded. Use when scanning who is on the network, spotting a new or unknown device, checking open ports, watching bandwidth or WAN throughput, auditing what a device phones home to, or baselining and diffing the network. Do not use for the router's own authoritative device list (use fritz), or filling and deploying Home Assistant automation templates (use homectl).

0
installs
▌
Pihole · larsboes bundle
Controls a Pi-hole v6 DNS resolver over its REST API — query stats, per-device egress (what each client phones home to), top domains/clients, pause or resume blocking, and add allow/deny rules. The Pi-hole host and admin password come from a Bitwarden item at runtime, never hardcoded. Use when the user wants Pi-hole DNS stats, per-device egress visibility, top domains or clients, blocked-query counts, ad-blocking pause/resume, or an allowlist/denylist entry. Do not use for router device listing (use fritz), Home Assistant entity control (use homectl), or non-Pi-hole DNS servers.

0
installs
▌
Secrets Guard · larsboes
Safe handling of secrets and credentials in a Claude Code session — how to use API keys, tokens, and .env / .secrets values for a task without ever pulling them into the model context, working WITH Axon's managed-settings deny policy instead of around it. Use when a task needs a credential (an authenticated API call, a deploy, a login), when you are about to read a .env / .secrets / credentials / .pem / .key file, or when a command was blocked by the security policy and you need the safe equivalent. Do not use for pi sessions (the secrets-guard extension enforces this there) or for creating/storing a new secret (that is tools/setup-secret.sh, run by the human).

0
installs
▌
Writing Skills · larsboes bundle
Authors, structures, and audits Claude agent skills (SKILL.md + scripts/references/assets) to Anthropic best-practices, the agentskills.io spec, and local Claude Code conventions. Use when creating a new skill, refactoring one for progressive disclosure, or auditing a skill's metadata/structure. Do not use for README/user docs, non-agentic library code, or MCP-server implementation — for a runbook-style domain skill, model it on a sibling under `.claude/skills/`.

0
installs
▌
Root Cause · larsboes bundle
Investigates why something already failed and names the systemic condition that allowed it, never the person who touched it last. Picks the method the evidence supports — timeline and change analysis, five whys, is/is-not, fault tree, blameless postmortem — and ends with a cause, the probe that would falsify it, and a fix ranked by whether it removes the whole class of failure. Use when something broke, keeps breaking, regressed, or behaved differently than expected and the question is why — an incident, a postmortem, an RCA, a recurring bug, "it worked yesterday", "this keeps happening". Do not use for attacking a proposal that has not been built yet (use red-team), for choosing between options (use council), or for scanning running software for vulnerabilities.

0
installs
▌
Esphome · larsboes bundle
Builds and OTA-flashes ESPHome firmware nodes through the esphome Docker container on a home-automation host — lists node YAMLs, compiles, uploads over WiFi, and streams device logs. Reaches the container over SSH (key auth) and the dashboard for a reachability check, with connection facts read from the committed overlay config at runtime, never hardcoded. Use when flashing an ESP32 or ESP8266, compiling an ESPHome node, running an OTA firmware update, tailing ESPHome logs, or adding a new sensor node. Do not use for Home Assistant automation-template materialization (use homectl), Fritz!Box network queries (use fritz), or general Docker container management.

0
installs
▌
Homectl · larsboes bundle
Materializes overlay-owned Home Assistant templates and vendors overlay-owned pinned components through a public, host-neutral CLI. Use for template coverage, materialization, or reviewed component vendoring. Real devices, automations, entity IDs, component selections, and output stay in the active private overlay.

0
installs
▌
Crystallize · larsboes bundle
Turn unstructured input — a raw brainstorm, a pile of notes, scattered ideas, a half-written PRD — into one precise document where every claim is grounded in the actual repo or vault and every decision is dated with its reasoning. Drives the process through batched multiple-choice questions rather than open-ended prompts, because the user has the judgment and you have the ability to measure. Use this whenever someone says "help me think this through", "make this precise", "turn this into a spec/PRD/plan", "I have a bunch of notes and no structure", "brainstorm with me", "ask me questions", "finalize this doc", or hands you a rambling document and asks what to do with it. Also use it when a document and the system it describes have drifted apart and you need to find out which is lying.

0
installs
▌
Ha Deploy · larsboes bundle
Validates, reloads, and verifies an existing Home Assistant deployment through its API. Use after overlay-owned configuration has been placed on the owning host. Container lifecycle and file deployment remain outside this skill.

0
installs
▌
Ha Dashgen · larsboes bundle
Generates Home Assistant Lovelace dashboards from an overlay-owned house model and deploys them, backing up first. Public mechanism, private inventory - this Pack owns every card, layout and deploy decision and knows no entity ID; the active overlay owns model.ts. Use when building, checking, or deploying generated dashboards, or when adding a room, entity or view to the generated set. Do not use for hand-editing a dashboard in the UI, backing up or restoring a single dashboard (use ha-dashboard), querying or controlling entities (use ha-cli), or filling automation templates (use homectl).

0
installs
▌
Home 3d Printing · larsboes bundle
Slices models and drives a home Klipper/Moonraker 3D printer via the local printctl CLI — slice with OrcaSlicer, upload, arm, start, monitor, pause/cancel, e-stop — with hard temperature caps and an arm-before-heat gate. Works with any Klipper/Moonraker printer; the specific model, material and caps come from config (run `printctl doctor`). Use when the user wants to 3D-print at home, slice an STL/3MF, or check/control the printer. Do not use for CAD/mesh modeling, cloud/non-local printers, or non-Klipper firmware.

0
installs
▌
Ha Dashboard · larsboes bundle
Deploys and backs up Home Assistant Lovelace dashboards over the WebSocket API — list existing dashboards, back up a dashboard's config to a file, push a config file to a dashboard, and create a new sidebar dashboard. Pure-Python-3 stdlib WebSocket client (no deps); the non-secret HA_URL comes from the committed overlay config and the secret HA_TOKEN from a Bitwarden item at runtime, never hardcoded. Use when deploying or backing up a Lovelace dashboard, saving or restoring dashboard config, listing dashboards, or creating a new dashboard. Do not use for querying or controlling entities or calling services (use ha-cli), router or network device listing (use fritz), DNS or ad-blocking (use pihole), or filling and deploying HA automation templates (use homectl).

0
installs
▌
Attention Control · larsboes bundle
Shape output for a reader with ADHD. Lead with action, use controlled vocabulary, active voice, and simple tenses. Invoke with /attention-control; stays on until "stop attention control".

0
installs
▌
Energy Dashboard · larsboes bundle
Configures the Home Assistant Energy Dashboard over the WebSocket API (energy/get_prefs, energy/save_prefs) — the REST API does not expose these prefs — wiring solar production/power, grid import/export, and battery charge/discharge sources, and reading the current sources back. Source entities are passed as flags or a raw JSON array, never hardcoded; existing grid price/cost config and the solar Forecast.Solar entry are preserved on write, so re-running is idempotent. The secret HA_TOKEN comes from a Bitwarden item at runtime and the non-secret HA_URL from the committed overlay config, never hardcoded. Use when wiring up or changing the Energy Dashboard sources, pointing the dashboard at solar/grid/battery sensors, or inspecting the current energy preferences. Do not use for general entity-state queries, service calls, or automation reloads (use ha-cli), or filling and deploying HA automation templates (use homectl).

0
installs
▌
Academic Writing · larsboes bundle
Drafts, critiques, and cite-checks academic writing (thesis chapters, conference/journal papers) across two genre profiles — empirical-CS/paper and DSR/qualitative-thesis — with a genuinely adversarial multi-skeptic critique pass and Quarto/BibTeX-aware citation tooling. Use when drafting a section from bullet points, running a flow/reverse-outline check on a draft, running a hard-critic or red-team review pass on a chapter or paper, or validating/resolving citations and checking citation-key coverage. Do not use for README/user docs, general non-academic prose, or as a ghostwriter — it produces sourced skeletons and critique, not net-new prose, unless the calling context explicitly grants an exception.

0
installs

29 published skills