t3chnaztea

Use when connecting to or operating a Batocera arcade/retro-gaming cabinet over SSH: establishing the connection, understanding its read-only rootfs and /userdata layout, finding config/log/ROM paths, applying changes safely, verifying a display/config change end-to-end without a controller, or diagnosing why a setting reverts to its default on every launch (configgen). Foundation for the other batocera-* skills (roms, display, tuning, maintenance). Covers sshpass, batocera.conf, EmulationStation, the ES HTTP API, and batocera-screenshot. Not for game recommendations.

Use when working with ROMs and gamelists on a Batocera cabinet: figuring out which directory a system's ROMs go in, why a game does not appear in the menu, which file formats and extensions a system accepts, curating a large library by hiding rather than deleting, editing gamelist.xml safely, cleaning up multi-disc games, or verifying dumps against No-Intro/Redump DATs. Covers extensionless-ROM-in-zip failures, the 3DO/Opera no-m3u trap, cave3rd, and the es_systems_custom.cfg landmine. Not for shaders/bezels (batocera-display), or ROM directories renamed by a Batocera version upgrade (batocera-maintenance).

Use when a Batocera cabinet feels slow, laggy, or stuttery, or when tuning performance: EmulationStation menu/carousel lag, a slow-since-boot frontend, CPU governor stuck in powersave, overclock or CPU-speed core options that seem to do nothing, run-ahead / input latency, framerate that drops the longer the machine runs, or a 3D-heavy game (PSX/PS2/Dreamcast/N64 class) that crashes or stutters — Vulkan errors in emulator logs, picking a rendering resolution or graphics backend the hardware can afford. Covers the diagnostic order (check the throttle first), the v43 governor-at-boot failure, the emulator-name vs rom-dir-name overclock scoping trap, ES event-hook frame cost, and hardware-tier detection (batocera-info / batocera-vulkan). Not for shader visual quality (batocera-display).

Use when working on how games look on a Batocera cabinet: CRT and other shaders, shaderset resolution and per-game shader overrides, bezel/overlay artwork and alignment, generating missing .info sidecars, filtering unusable bezels, pinning the display resolution, or diagnosing moire and cropped/misaligned game viewports inside bezels. Covers why a shaderset seems invalid, why .info top values are not alpha coordinates, and why Mega-Bezel per-game art fails. Not for performance/framerate tuning (batocera-tuning).

Use when keeping a Batocera cabinet healthy over time: running scripts or scheduled jobs at boot and on an interval without cron, the custom_service boot-service pattern (which replaced custom.sh in v43), backing up /userdata to another machine over SSH, and upgrading between major Batocera versions (v42 to v43 and similar) including the ROM-directory renames and post-upgrade checklist. Covers why there is no crond, the backup-before-upgrade doctrine, and catch-up scheduling. Not for performance tuning (batocera-tuning).

Use when [the specific situations, symptoms, and phrases that should trigger this skill]. Pack the words someone would actually type or think when they hit this problem. Describe ONLY when to use it, never summarize the workflow inside. Keep the whole frontmatter under 1024 characters. End by naming what this skill is NOT for, pointing at the sibling skill that covers it.

Use when connecting a coding agent to a Home Assistant instance for the first time or in a fresh session: "connect to Home Assistant", "SSH into HA", set up the Terminal & SSH add-on, generate a long-lived access token, call the REST API, check an entity's state, call a service from the command line, edit configuration or automations YAML over SSH, figure out which of the two access lanes (SSH vs REST) a task needs, or decide whether an MCP integration already covers the task instead. Foundation skill: the other ha-* skills assume its connection pattern and doctrine. Not for writing automations (ha-automations) or building the entity inventory (ha-context-map).

Use when writing, editing, debugging, or verifying Home Assistant automations with an agent: "write an automation that...", turn an English description into automation YAML, edit automations.yaml over SSH, an automation that never fires or fired once and went dead, a service call that silently did nothing, testing with automation.trigger, or reloading automations without restarting Core. Covers the write → check → reload → verify loop and the silent-failure traps (event-entity triggers, dead notify services, no-op writes). Not for the initial connection (ha-connect) or price/LLM-driven triggers (ha-external-triggers).

Use when an agent keeps guessing wrong entity ids, when starting recurring agent work against a Home Assistant instance, or when asked to "map my Home Assistant", "build an entity inventory", "document my smart home", or "why does the agent keep using entities that don't exist". Covers writing and maintaining the instance map file: entity inventory, add-ons, integrations, helpers, notify targets, and a running Gotchas list, plus the provenance checks that keep it honest. Not for making the initial connection (ha-connect) or writing automations (ha-automations).

Use when a Home Assistant automation's "if" is not a device: react to a utility's real-time electricity price (load shedding, price-spike alerts, demand response), or use an LLM as a sensor via ai_task and a camera snapshot for judgment calls (person vs cat vs shadow, "is the garage door actually open in this image"). Covers the save/bump/shed/restore thermostat loop with hysteresis, threshold helpers, the armed-sentinel restore guard, ai_task.generate_data with structured output and camera attachments, and branching automations on LLM fields. Not for basic automation mechanics (ha-automations) or connecting (ha-connect).

Use when [the specific situations, symptoms, and phrases that should trigger this skill]. Pack the words someone would actually type or think when they hit this problem. Describe ONLY when to use it, never summarize the workflow inside. Keep the whole frontmatter under 1024 characters. End by naming what this skill is NOT for, pointing at the sibling skill that covers it.

Use when deciding what to do with started-but-unfinished PlayStation games: "triage my backlog", "what should I go back and finish", "what should I drop", "which games am I never finishing", "clear my pile of shame", "should I give up on X". Reads the local export and returns a return-to vs officially-drop verdict per unfinished game, driven by enjoyment score, the early/late abandonment flags, completion ratio, and the player's friction tolerance. Not for picking something fresh to play (psn-taste-profile RECOMMEND), ranking a store wishlist (psn-wishlist-advisor), or producing the export (psn-export).

Use when reading a PlayStation player's taste from their local export: "what does my PSN library say about me", "read my gaming taste", "what should I play next", "recommend a game from my library", "analyze my play style", "did my taste change", "compare my profile to last month". Reads preferences.json (the taste profile psnstats produces) and turns it into a palette read plus recommendations that cite the specific signal behind each pick. Also owns --compare drift interpretation (how taste shifted between two exports). Not for return-vs-drop verdicts on unfinished games (psn-backlog-triage), ranking a store wishlist (psn-wishlist-advisor), or producing the export itself (psn-export).

Use when deciding what to actually buy off a PlayStation Store wishlist: "rank my wishlist", "what should I buy next", "is this sale worth it", "what on my wishlist will I actually finish", "clean up my wishlist", "should I grab X while it's discounted". Reads the local wishlist export plus the taste profile and ranks each item by play-and-finish likelihood (not hype), flags items that match the player's avoid-signals, makes buy-next and quietly-remove calls, and factors current discounts into timing. Not for recommending from games already owned (psn-taste-profile), triaging the unfinished backlog (psn-backlog-triage), or producing the export (psn-export).

Use when [the specific situations, symptoms, and phrases that should trigger this skill]. Pack the words someone would actually type when they want this. Describe ONLY when to use it, never summarize the workflow inside. Keep the whole frontmatter under 1024 characters. End by naming what this skill is NOT for, pointing at the sibling skill that covers it.