Claude Skills by mr-tbot

Use when the user wants subagents spawned economically — they say \"/auto-agent-eco\" (optionally followed by a number capping how many subagents), ask to keep agent costs down, to orchestrate cheaply, to use cheaper models for the legwork, or to stop defaulting every subagent to the heaviest model. Sets the model-tier policy for a fan-out: the orchestrator stays top tier and every subagent runs strictly below it.

Use when the user wants the ceiling off subagent spawning — they say \"/auto-agent-max\" (optionally followed by a number capping how many subagents), ask for maximum capability, peer-tier subagents, no model downgrade, or the best possible result regardless of cost. Removes the tier cap so the orchestrator may spawn same-tier subagents, while still sizing each one to the task.

Use when a codebase needs an adversarial security review or compliance readiness assessment — web UI and API security, authentication and access control, licensing or paywall bypass, secrets exposure, dependency vulnerabilities, and SOC 2 / ISO 27001 / HIPAA / PCI DSS / GDPR readiness; when the user mentions security audit, pen test, red team, blue team, purple team, white team, tabletop or adversary emulation, SOC 2, compliance, Vanta, hardening, or says \"/auto-audit-security\"; or when abo...

Use when a project must be driven to genuine completion on every platform it targets — user says \"finish this\", \"don't stop until it's done\", \"100% complete\", \"keep doing passes\", \"audit this until it's done\", \"/auto-audit\"; when they ask for a red-team, blue-team or multi-team pass over a project's correctness; or when about to declare a feature done, a capability impossible, a platform unsupported, or a build ready to ship.

Use at the start of a session on a machine running several editors, agents or builds at once — when builds fight for CPU/RAM, an OOM kill lands, devices are contested (Android phones, ESP32 or dev boards on serial, iOS devices on a remote Mac), or the user mentions resource limits, parallel sessions, subagent count, account limits, or says \"/auto-balance\"; or before starting a long build or flashing a board. Resource admission itself is handled by `aw` (agent-awareness); this skill covers w...

Use when a product's brand must be consistent everywhere it appears — app icons across Windows, macOS, Linux, iOS, Android and the web, favicons and PWA manifests, store listings, splash screens, logos, colors and fonts; when the user mentions branding, icons, logo, favicon, app icon, brand assets, or says \"/auto-brand-parity\"; or when about to generate an icon set or call branding consistent without comparing platforms.

Use when a codebase's comments need bringing into line with the code — inconsistent commenting across files written by different hands or models, comments that describe behavior that changed, missing doc comments on public API, or over-commented noise; when the user says fix the comments, document the code, add docstrings, or \"/auto-comment\"; or when about to leave a changed function with a comment that no longer describes it.

Use at the start of every session, and before any command that touches physical hardware — adb, fastboot, scrcpy, screenshots, screen recording, esptool/idf.py/platformio, a serial console, xcrun devicectl — on a machine where more than one editor window or agent session may be running; when phones, dev boards or test devices are shared between projects; when a device behaves as though something else is driving it; when a lock looks stale after a crash; or when the user says \"/auto-device-lo...

Use when documentation must be brought back in line with the code — README, CHANGELOG, docs sites, API docs, wikis, GitHub/Forgejo repo surfaces, and external project-management tools; when the user asks to update docs, check whether docs are stale, verify the changelog, audit feature claims, or says \"/auto-doc\"; or before a release, when shipping a UI or API change, or when about to call documentation current without checking.

Use when a project should be taken from wherever it is to genuinely shippable in one pass — issues triaged and fixed, code audited, UI and docs and media and infrastructure caught up; when the user says do everything, full pass, get this release-ready, run all the audits, or \"/auto-everything\"; or when several of the Auto-* skills would each apply and running them in the right order matters.

Use when reported problems need to be found, fixed and answered — GitHub/Forgejo issues, Jira, Linear, ClickUp, Trello, Zendesk, Slack or Discord reports, and crash reporters like Sentry, Crashlytics and Play vitals; when the user says triage the issues, fix reported bugs, respond to issues, clear the backlog, or \"/auto-issue-fix\"; or when about to comment publicly, close someone's issue, or mark a crash resolved.

Use when a codebase's licensing must be audited against how it will actually be released — before shipping, selling, open-sourcing, fundraising, or an acquisition; when the user asks about dependency licenses, GPL/AGPL/copyleft exposure, whether they can go closed-source, what license to pick, NOTICE or attribution files, \"/auto-license-check\"; or when about to declare a dependency's license fine, unknown, or unfixable.

Use when a project needs product video made or kept current — onboarding, tutorials, walkthroughs, launch stingers, social cutdowns, release-notes clips; when the user mentions HeyGen, avatar video, demo video, tutorial video, product ads, or says \"/auto-media-maker\"; or when code has changed and the videos that teach it may now be stale.

Use when a project's first-run onboarding video — the one a new user watches before they know what anything does — needs making or bringing back in line with the current first run; when the user mentions the onboarding video, the welcome video, the getting-started clip, first-run, or the walkthrough new users see, or says \"/auto-media-onboarding\"; or when signup, install, permissions or the first screen has changed and the video still teaches the old path.

Use when a project's main stinger — the short launch commercial that plays before anything else — needs writing, cutting, or bringing back in line with what the product now is; when the user mentions the stinger, the launch ad, the trailer, the sizzle reel, the hype cut, the thirty-second spot, or says \"/auto-media-stinger\"; or when the name, price, tagline, platforms or hero screen has moved and the ad still sells the old product.

Use when a codebase must be provably the user's own work — before a release, sale, funding round, acquisition, or open-sourcing; when code may have been copied from another project, Stack Overflow, or regurgitated by an AI; when the user says \"check for plagiarism\", \"is this code original\", \"clean-room\", \"copyright risk\", \"/auto-rewrite\"; or when about to declare copied code fine, or impossible to replace.

Use when the agent skills, plugins and marketplaces installed on this machine may have fallen behind the repositories they came from — when the user asks whether their skills are current, wants them updated, mentions a skill behaving like an older version, has just set up a new machine, or says \"/auto-skill-update\"; or before relying on a skill whose upstream has moved.

Use when a project's UI needs to be brought to a professional finished state — when it was built in phases by different people or models and now has competing design idioms, when screens are inconsistent or half-finished, when the user asks to audit/polish/unify/fix the UI or UX, mentions design cohesion, accessibility, contrast, spacing, dark mode, buttons or controls that do nothing, or says \"/auto-ui-ux\"; or when about to declare a UI done without looking at it.

Use when deployed web or backend infrastructure must be checked against the repository — when the user asks whether prod matches main, mentions deployment drift, infra drift, TLS or DNS or CDN or env-var parity, asks to verify a deploy, or says \"/auto-web\"; or when about to claim a deployment is current, an environment is configured correctly, or infrastructure state is safe to change.