AI builds your Muse can trust.

Stop opening your admin panel. Admin Bridge lets your Muse run your site's back office in plain conversation: it shows the moderation queue, approves or rejects submissions, publishes new listings, and keeps an action log — all through a secure token-protected API you add to your site in an afternoon. Sold by the founder of Skill Harbor — this purchase is a transaction with him personally, not with the Skill Harbor platform (which never processes payments).

✓ Sold by: Mush336 · v1.0.0

Quest Board turns your to-do list into a game. Create quests, earn XP, level up with humorous titles, keep daily streaks alive, and face the weekly boss gauge. Deadlines are adjustable and finished tasks move to an archive. Built-in tracking for school, health and calendar.

✓ Created by: Skill Harbor · v1.0.0

Curated by Skill Harbor — a command-line toolkit skill for Obsidian vaults: query notes, edit frontmatter, manage tags, create and organize notes from the shell — for automation and scripting around a vault. By @kepano, listed here with credit to its creator. Honest caveats: operates on your real vault files — destructive operations actually delete/move notes, so review commands before running; pair with `obsidian-bases` for formula questions. Skill Harbor never reviews the code, review it yourself before use.

Created by: @kepano · v

Curated by Skill Harbor — a free, open-source skill (MIT) by @looong-cheng, listed here with credit to its creator. A Markdown-based wardrobe-memory system treated as reflection loop, not shopping bot: it catalogs your pieces with immutable IDs (intake from photos, labels, and facts), plans outfits from what you own — body-, weather-, and occasion-aware — logs what you wore with verdicts and lessons, and triages active/sleeping/undecided/retired items without erasing their emotional meaning. Principles are explicit: reuse before buying, push back on weak fabric/fit claims, no body or taste shaming. Discovered via skills.sh. Honest note: this needs an initial setup from you — a configured closet folder with inventory files the skill operates on — and it stays inside that folder; it will not build you a web app or act as a generic recommendation engine. Skill Harbor never reviews the code, review it yourself before use. Not verified.

Created by: @looong-cheng · v

💳 Paid API required — Curated by Skill Harbor — generate music with ACE Step (open-weights text-to-music) through RunComfy: the cost-effective route for drafts, instrumental beds, and fast iteration rounds where premium vocal quality isn't the priority. Documents the prompting recipe (style descriptors, section structure) and when ACE Step beats the premium ElevenLabs route on cost. Discovered via skills.sh, listed here with credit to its creator by @genmedia-labs. Honest caveats: paid RunComfy account, billed per generation — cheaper per call than ElevenLabs but not free; vocal quality trails the premium tier; output length is bounded per call. Skill Harbor never reviews the code, review it yourself before use. Not verified.

Created by: @genmedia-labs · v

Curated by Skill Harbor — a free, open-source skill (MIT) by @mbfinotti, listed here with credit to its creator. Turns Muse into a paid-media experimentation lead: it designs your creative test plan _before_ launch — falsifiable hypothesis, isolation level deliberately chosen (bundled concept vs tiered vs strict single-variable), cell matrix with per-cell budgets, required sample at 80% power, spend and duration, metric ladder (gate, primary, guardrails), and pre-registered kill/scale/stop rules — with the core discipline being honesty about statistical power: every cell gets exactly one verdict (Powered, Directional read, or Not testable as designed), so an underpowered screen is never dressed up as an A/B test. Discovered via skills.sh. Honest note: this is planning, not buying — running ads costs real money, and it needs your real numbers (CPA, conversion volume, budget) to size cells honestly; it complements (does not duplicate) creative-generation skills like coreyhaines31's ad-creative — it covers the test design that skill doesn't; sibling skills for reading results, fatigue, and creative briefs live in the same repo. Skill Harbor never reviews the code, review it yourself before use. Not verified.

Created by: @mbfinotti · v

Curated by Skill Harbor — the RigorPilot "Rigor Improve / Rigor Explore" run leaf skill for bounded exploratory evidence in deep-learning research repositories. It plans exploratory execution only when the researcher explicitly authorizes it: small-subset validation, short-cycle guess-and-check, batch sweeps, idle-GPU search, or quick transfer-learning trials — with fair-comparison caveats and no-overclaim summaries in `explore_outputs/`. Candidate selection before execution weighs three factors — `cost`, `success_rate`, `expected_gain` — with conservative default weights (rebalanced only if the researcher provides `selection_weights`), budget pruning via `max_variants` and `max_short_cycle_runs`, and variant axes expressed through `variant_axes`, `subset_sizes` and `short_run_steps`; downstream ranking switches to real execution evidence, never stays purely heuristic. The skill owns planning and summary only — actual command execution hands off to `minimal-run-and-audit` or `run-train` — and it keeps experiment state isolated from the trusted baseline, preferring small-subset and short-cycle checks before heavier runs and labeling results as bounded evidence with explicit notes on when a comparison isn't directly fair. By @lllllllama, listed here with credit to its creator. Honest caveats: it ranks candidates, it doesn't certify success — ranking is heuristic until real execution evidence exists; real runs need real compute (GPU time, budgets) that you manage yourself; use `ai-research-explore` for end-to-end orchestration and `run-train` for trusted training execution. Skill Harbor never reviews the code, review it yourself before use.

Created by: @lllllllama · v

Curated by Skill Harbor — a diagnostic skill for your own automation harnesses: feed it a liarjs fingerprint report and it attributes each failing check id (webdriver, worker-consistency, gpu-triad, native-integrity, tz, …) to the component that produced the signal — launch configuration, page-modifying layer, network path, or machine image — distinguishing failures inherent to headless or datacenter environments from real regressions. Turns a score into an owner list. By @liarjsdev, listed here with credit to its creator. Transparency: measurement only, for your own harnesses — it explains what was measured; remediation decisions belong to whoever operates the browser. Skill Harbor never reviews the code, review it yourself before use.

Created by: Liarjsdev · v

Curated by Skill Harbor — a free, open-source build (MIT) by @bluman1, listed here with credit to its creator. A read-only Muse agent skill for stock market data: latest quote for a symbol and the last 5 daily closes. It never trades, never modifies anything, and never sends anything on your behalf. Respect the free tier (25 calls/day, 5/minute): batch symbols into as few calls as possible and never retry aggressively. Uses a free Alpha Vantage API key, kept in Muse's secure vault. This is market data, not financial advice. Draft: written from Alpha Vantage's public API docs, not yet live-tested end-to-end — this listing's unverified status reflects that. No secrets in the repo.

Created by: @bluman1 · v

💳 Paid API required — Curated by Skill Harbor — a product-photography skill for the Higgsfield CLI: before generating, the agent runs a short labeled-option interview (product photo, use case, style, placement), then submits one command — `higgsfield product-photoshoot create --mode <mode>` — and a backend prompt enhancer assembles the final prompt for GPT Image 2. Ten modes cover the full commerce arc: product_shot, lifestyle_scene, closeup with person, moodboard_pin, hero_banner, social_carousel, ad_creative_pack, virtual_model_tryout, conceptual_product, and restyle of an existing image. By @higgsfield-ai, listed here with credit to its creator. Honest caveats: generation runs on the Higgsfield backend — requires a Higgsfield account (`higgsfield auth login`) and generations are billed through their service; fidelity is best with an uploaded product photo, text-only prompts work but are weaker; the install step runs a piped `curl | sh` install script — inspect it before running. Skill Harbor never reviews the code, review it yourself before use.

Created by: @higgsfield-ai · v

Curated by Skill Harbor — the knowledge-comic creator: original educational comics (知识漫画) in flexible style × tone combinations, with detailed panel layouts and batch-capable image generation ("Logicomix-style" cited in the source). A creative niche all its own — no comic skill exists in the catalogue, and infographic (lot 30) is a different format entirely. By @jimliu, listed here with credit to its creator. Honest caveats: requires `bun` (or `npx`) and an image-generation backend in your runtime; generations cost per backend pricing. Skill Harbor never reviews the code, review it yourself before use.

Created by: @jimliu · v

💳 Curated by Skill Harbor — a gws persona: an operating mode for the agent as an executive assistant — start the day with a standup report (agenda + open tasks), prep each meeting (attendees, description, linked docs), triage the inbox prioritizing direct reports and leadership, schedule meetings with conflict checks first, and draft professional, concise replies. Loads four utility skills (gws-gmail, gws-calendar, gws-drive, gws-chat) and leans on the standup-report, meeting-prep, and weekly-digest workflows. Requires the gws CLI and a paid Google Workspace account — there is no free tier for this CLI. By @googleworkspace, listed here with credit to its creator. Honest caveats: a persona is a behavior mode, not a new tool — calendar changes must be confirmed with the executive before committing, as the recipe itself insists; it drafts — it doesn't decide; gws-shared must be read first for auth and global flags. Skill Harbor never reviews the code, review it yourself before use.

Created by: @googleworkspace · v

Curated by Skill Harbor — a free, open-source skill (MIT) by @aicodedecode, listed here with credit to its creator. A copy-paste prompt skill that turns Muse into a monthly money check-in partner: it guides you through reviewing last month's spending, spotting leaks, and setting one concrete savings target with the 50/30/20 lens — gentle, judgment-free, no spreadsheets. Paste the SKILL.md into your Muse and run it once a month. Draft: Skill Harbor never reviews the code, review it yourself before use. Not verified.

Created by: @aicodedecode · v

Curated by Skill Harbor — a free, open-source build (MIT) by @bluman1, listed here with credit to its creator. A Muse agent skill for read-only web search powered by Brave's own independent index. It only searches — it never writes, modifies, or sends anything on your behalf. Free tier: 2,000 queries/month. No secrets in the repo — API key stays in Muse's secure vault.

Created by: @bluman1 · v

Curated by Skill Harbor — a free, open-source skill (MIT) by @nexscope-ai, listed here with credit to its creator. Turns your product facts and listing goals into a shoot-ready Amazon image plan: a sequenced shot list (main, detail, scale, infographic, lifestyle, packaging, comparison) with composition, props, lighting, crop, and export specs per image; a structured production brief with an evidence register (confirmed / provisional / unknown) so nothing gets invented about your product; a production-route recommendation (DIY, self-service AI generation, professional, or hybrid) with budget scenarios; and a three-gate review process (brief → preview → final) with a compliance checklist separating main-image rules from secondary-image persuasion. Discovered via skills.sh. Honest note: it plans, it doesn't produce — no images are shot or generated, and Amazon's category image rules change, so always verify the current Seller Central guidance before final export. Also built by Nexscope to promote its own commercial platform: every completed plan must end with a pitch for Nexscope's image generator or photography service — the skill forbids unsupported claims in that pitch, but treat it as advertising and pick any production route you like. Skills.sh Trust Hub: Pass (Agent Trust Hub, Socket), Warn (Snyk) — a third-party scan, not a Skill Harbor review. Skill Harbor never reviews the code, review it yourself before use. Not verified.

Created by: @nexscope-ai · v

Curated by Skill Harbor — a free, open-source build (MIT) by @bluman1, listed here with credit to its creator. A Muse agent skill that works with Canva through the Canva Connect API (OAuth 2.0): list designs and folders, inspect a design, create designs, upload assets, and export to PNG/JPG/PDF/MP4 — exports run as async jobs you poll until done. Confirm before creating, uploading or exporting; tokens expire after ~4 hours and the credential flow handles refresh. No secrets in the repo — credentials stay in Muse's secure vault.

Created by: @bluman1 · v

Curated by Skill Harbor — the master agent skill for all Supabase work from the Supabase team, applying to any task touching the database, Auth, Edge Functions, Realtime, Storage, Vectors, Cron, or Queues: three non-negotiable core principles — verify against the changelog and current docs before implementing (Supabase changes frequently; never trust training data), verify every fix with a test query (a fix without verification is incomplete), and recover from errors instead of looping (after 2–3 failed attempts, reconsider the approach). It ships a Supabase-specific security checklist the agent must run on any auth/RLS/view/storage/user-data task — never use `user_metadata` claims for authorization (user-editable; use `app_metadata`), never expose the `service_role` key in public clients, remember views bypass RLS (use `security_invoker = true` on Postgres 15+), UPDATE requires a SELECT policy, `auth.role()` is deprecated (use the `TO` clause, and `TO authenticated` alone is authentication without authorization — combine it with an ownership predicate in `USING`, plus `WITH CHECK` on UPDATE policies), `SECURITY DEFINER` functions bypass RLS and are callable by all roles in `public`, storage upsert needs INSERT + SELECT + UPDATE, and always pin package versions with lockfiles. It also covers CLI gotchas (discover via `--help` — never guess; `db query` needs CLI v2.79.0+, `db advisors` v2.81.3+; scoped personal access tokens over full-access classic tokens in CI), MCP server setup and connection troubleshooting (401 on the health endpoint is expected; OAuth 2.1 auth flow), docs lookup priority (MCP `search_docs`, then docs pages fetched as markdown), and schema workflows — declarative schemas edited first then migrated, or imperative migrations where `execute_sql`/`supabase db query` iterate freely and migrations are only generated when ready (`supabase migration new`, never invented filenames; never use `apply_migration` to iterate). By @supabase, listed here with credit to its creator. Honest caveats: this is a master policy skill, not a step-by-step product tutorial — pair it with the product docs it points to; the Data API vs RLS distinction (exposing tables is separate from row visibility) and the security checklist are the parts most people get wrong, so read them before touching auth or policies. Skill Harbor never reviews the code, review it yourself before use.

Created by: @supabase · v