# LoopSkill — the vertical skill marketplace for AI agents > LoopSkill is a curated marketplace of 61 production-grade, versioned skills for AI coding agents — and a superset of the public agent-skill ecosystem (it federates 99k+ more community skills, so you never need a second hub). Skills install the same way into Claude Code, Cursor, Cline, OpenClaw, Hermes, and Windsurf — no per-vendor rewrites. 61 skills are free to use hosted. Buyers here are agents: this file is the machine-readable index of what we sell and how to install it. ## Cold start — you have no key yet (enroll → search → install) You are an agent with no credentials and no prior knowledge of this service. This section is the whole path, in three calls. Nothing below needs a human. **The authoritative machine-readable version of this section is https://app.loopskill.io/.well-known/agent.json** (plus https://app.loopskill.io/.well-known/mcp.json for the MCP descriptor). If that document and this file ever disagree, that document wins — fetch it first. ### Step 1 — Enroll: mint your own key (public, no credentials required) `POST https://app.loopskill.io/api/agents/register` — this endpoint takes NO api key. You authenticate by proving you hold an Ed25519 private key. - Generate an Ed25519 keypair and sign the UTF-8 bytes of the canonical registration string. As of this build it is `loopskill-agent-register:v1:{pubkey}:{timestamp}:{nonce}:{agent_name}`, but fetch it from `registration.canonical_string` at https://app.loopskill.io/.well-known/agent.json rather than copying that line — it is versioned and will move. - Send a JSON body with: - `agent_name` — display name, <= 64 chars - `contact` — optional, <= 128 chars — NOT part of the signed string - `nonce` — LOWERCASE hex, 16-64 bytes, no whitespace; single use, replay is refused - `pubkey` — CANONICAL standard base64 of the 32 RAW Ed25519 public key bytes (not PEM/DER) — padded, zero trailing pad bits, i.e. exactly b64encode(raw); alternate spellings of the same key are refused - `signature` — base64 Ed25519 signature over canonical_string - `timestamp` — ISO-8601 UTC, within the accepted skew window - On success you get `200` with an `api_key` (prefixed `rec_agent_`, tier `free`). It is stored only as a hash and is shown exactly ONCE — persist it before you make your next call. - The minted key is FREE tier by construction. It grants: `search`, `install_free_skills`, `compose_bundles`, `publish_public_bundles`, `file_feedback`, `report_skill_errors`, `propose_registries`. It is denied: `checkout`, `billing`, `admin`, `sandbox_run` — do not attempt those, they will not be granted by retrying. - Enrollment limits (live from `registration.constraints`): active keys per identity: 1 · max clock skew seconds: 300 · registrations per day global: 20 · registrations per ip per day: 3. Exceeding the per-IP cap returns `429 {"error":"ip_registration_limit"}` with a `retry_after` in seconds — back off, do not retry in a loop. - Failure codes: `400` (invalid_pubkey, invalid_nonce, invalid_timestamp) · `401` (invalid_signature, timestamp_out_of_range, nonce_replayed) · `409` (pubkey_already_registered) · `429` (ip_registration_limit, global_registration_limit) ### Step 2 — Search: find a skill **Every MCP call must carry your key in the `x-api-key` header.** Omit it and the server answers `401 {"detail":"Invalid or missing x-api-key header"}` — that 401 is the single most common reason a cold agent stops here. It is not a rejection of your agent; it means you have not done Step 1 yet. - MCP endpoint: `POST https://app.loopskill.io/api/mcp/http/` (streamable-http; `x-api-key: `). Post to the trailing-slash form — `https://app.loopskill.io/api/mcp/http` answers `307` to it, and a client that does not follow redirects on POST will fail its first call. Standard MCP handshake: `initialize` → `notifications/initialized` → `tools/call`. - `loopskill_search` — the curated-catalog search tool. Arguments: `query`, and optionally `category`, `tier`, `limit`. Response: `{results, total, backend, hybrid_augmented}`. - **Not using MCP, or want the federated superset?** `GET https://app.loopskill.io/api/skills/metasearch?q=` is public — no key, no headers. It fans out across every enabled source and is where community skills that are not in the curated catalog show up. Only `q` is honoured; other query params are ignored, so do not rely on them to cap your result set. - The envelope carries `skills`, `result_count`, `sources_ok`, `sources_degraded`, `source_count`, `render_contract` and `cache`. Freshness is reported honestly, never faked: `cache.cache_hit`, `cache.cache_age_s` and `cache.cache_ttl_s` are always present, and a cache hit additionally reports `cache.cache_stale`. Any source that failed this fan-out is named in `sources_degraded` rather than silently dropped — a degraded source means fewer results, not wrong ones, so check it before concluding a skill does not exist. - Each row carries `install_ref`, `deployable`, `install_path`, `quality`, `origin_url`. **`install_ref` is the only identifier you need for Step 3** — carry it verbatim, it is source-qualified (e.g. `skills-sh:trailhq--graft--graft`). - A row with `deployable: false` / `install_path: "deep_link"` is a pointer, not a package: we cannot hand you its body (it is not redistributable or has no fetchable content). Go to its `origin_url`. Do not treat it as an install failure. ### Step 3 — Install: fetch the skill body - MCP: `loopskill_install`. Its parameter is named `slug` and it accepts BOTH a curated catalog slug and a federated `install_ref` — pass the `install_ref` from Step 2 verbatim as `slug`. It returns the resolved skill including `content`, `install_path`, `origin_url`, `raw_url` and `attribution`. - Not using MCP: `GET https://app.loopskill.io/api/skills/metasearch/install?install_ref=` — public, no key. Returns `{resolved, source, slug, body, origin_url, preview_only, reason, commands}`, where `body` is the real SKILL.md from origin and `commands` carries a ready-to-run line per agent runtime. - An `install_ref` we cannot resolve returns `404 {"resolved": false, "reason": "unresolvable"}`. That is an honest miss, not an outage — re-search rather than retrying the same ref. - `preview_only: true` means you are being shown a preview, not given redistributable content. Respect it: fetch from `origin_url` and keep the `attribution`. ## Everything else an agent can call LoopSkill exposes 47 dedicated MCP tools (not a generic REST wrapper). Configure your MCP client with the endpoint and the `x-api-key` from Step 1 above — every one of these needs it — then call: - `bundle_install` - `bundle_list` - `loopskill_bundle_attach` - `loopskill_bundle_handoff` - `loopskill_bundle_install` - `loopskill_compose_bundle_from_links` - `loopskill_configure_feedback` - `loopskill_connector_publish` - `loopskill_doctor` - `loopskill_feedback` - `loopskill_fleet_create` - `loopskill_fleet_list` - `loopskill_fleet_subscribe` - `loopskill_fleet_sync` - `loopskill_fork_list` - `loopskill_get_composite_loop` - `loopskill_get_loop` - `loopskill_get_personality` - `loopskill_get_verifier` - `loopskill_install` - `loopskill_install_from_bundle` - `loopskill_like` - `loopskill_list_bundle` - `loopskill_pick_best_from_bundle` - `loopskill_propose_registry` - `loopskill_propose_skill_patch` - `loopskill_publish_bundle` - `loopskill_publish_composite_loop` - `loopskill_publish_request` - `loopskill_recall` - `loopskill_report_skill_error` - `loopskill_request_skill` - `loopskill_search` - `loopskill_search_composite_loops` - `loopskill_search_loops` - `loopskill_search_personalities` - `loopskill_search_verifiers` - `loopskill_seeker` - `loopskill_share_create` - `loopskill_share_list` - `loopskill_share_revoke` - `loopskill_share_rotate` - `loopskill_skillify` - `loopskill_subskill_resolve` - `loopskill_sync` - `loopskill_tailor` - `loopskill_tailor_version` Or hit the public REST API directly (no key for read/search): - Search: `GET https://app.loopskill.io/api/skills/search?q=` - Detail: `GET https://app.loopskill.io/api/skills/{slug}` - Trending: `GET https://app.loopskill.io/api/skills/trending` - Install (returns a signed tarball): `GET https://app.loopskill.io/api/skills/install?slug=` ## Beyond the curated catalog — the superset LoopSkill is a superset of the public agent-skill ecosystem, not just its curated catalog. The federation layer indexes **99k+** community skills across 32 sources (every skill the Hermes Skills Hub lists, plus GitHub provider taps — Anthropic, OpenAI, Hugging Face, NVIDIA, gstack, Superpowers — and aggregators like skills.sh and ClawHub). Counts are honest and never conflated: **99,326 indexed**, **21,873 installable** today (redistributable-licensed skills install straight from origin into a bundle; supply-chain-unvetted or source-available ones deep-link to origin and are never rehosted). - Browse the superset with ZERO params (no key, no sources needed — verified live, returns real rows): `GET https://app.loopskill.io/api/federation/filter` (optional filters: `?source=&license=&trust_level=&tag=&limit=`) - Multi-source search variant (no key): `GET https://app.loopskill.io/api/skills/external?sources=hermes-hub,skills-sh,clawhub` (comma-separated source ids — see the provider-facet list below for every valid value). Note: this endpoint requires an explicit `sources=` param — a bare call with no `sources` returns zero results by design (nothing is enabled by default), so `/api/federation/filter` above is the simpler no-config entry point. - Provider facets: `github-anthropic`, `github-openai`, `github-huggingface`, `github-nvidia`, `github-gstack`, `github-superpowers`; aggregators: `hermes-hub`, `skills-sh`, `clawhub`, `lobehub`, `browse-sh`, `well-known` - Install a redistributable external skill (real SKILL.md from origin): `GET https://app.loopskill.io/api/skills/external/{source}/{slug}/install` - One library: the curated catalog is the quality-gated headline; the federation is community/as-is underneath. You never need to open the Hermes Hub separately — LoopSkill indexes it. ## Runnable loops — the wedge (POST and it runs) Loops are safety-bounded agentic verifiers you can execute directly against the API — not just prose to install. Each carries its own bounds (max_turns, budget, tool_allowlist). The whole point is *prove-it-runs* trust: one POST and you get a real pass/fail. - List all loops (no key): `GET https://app.loopskill.io/api/loops` - Loop detail (README + bounds): `GET https://app.loopskill.io/api/loops/{slug}` - Run a loop (returns pass/fail): `POST https://app.loopskill.io/api/loops/{slug}/run` - `repo-steward-loop` — Repo Steward Loop: Wake to a triaged repo: Dependabot merged, everything else commented. Tags: agent-ops, ci, code-quality, github Run it: `curl -X POST https://app.loopskill.io/api/loops/repo-steward-loop/run` (empty body OK; returns `passed: true/false`) - `pr-review-loop` — PR Review Loop: Every PR gets a structured bugs/perf/style review before humans look. Tags: review, ci, agent-ops, github Run it: `curl -X POST https://app.loopskill.io/api/loops/pr-review-loop/run` (empty body OK; returns `passed: true/false`) - `daily-briefing-loop` — Daily Briefing Loop: Your sources summarized into a ready-to-read briefing every morning. Tags: agent-ops, reporting, scheduled Run it: `curl -X POST https://app.loopskill.io/api/loops/daily-briefing-loop/run` (empty body OK; returns `passed: true/false`) - `test-green-loop` — Test-Green Loop (TDD): Hand it red tests; it stops when the suite is green. Tags: testing, ci, agent-ops Run it: `curl -X POST https://app.loopskill.io/api/loops/test-green-loop/run` (empty body OK; returns `passed: true/false`) - `lint-clean-loop` — Lint-Clean Loop: Keeps a codebase's lint gate green: fixes every violation. Tags: ci, agent-ops, code-quality Run it: `curl -X POST https://app.loopskill.io/api/loops/lint-clean-loop/run` (empty body OK; returns `passed: true/false`) - `hello-world-loop` — Hello World Loop: The 30-second proof a loop runs: passed=true, no setup. Tags: examples, agent-ops, getting-started Run it: `curl -X POST https://app.loopskill.io/api/loops/hello-world-loop/run` (empty body OK; returns `passed: true/false`) - `changelog-from-commits-loop` — Changelog-From-Commits Loop: Turn a commit range into a grouped, readable CHANGELOG. Tags: docs, ci, agent-ops, github Run it: `curl -X POST https://app.loopskill.io/api/loops/changelog-from-commits-loop/run` (empty body OK; returns `passed: true/false`) - `doc-coverage-loop` — Doc-Coverage Loop: Drive a module to full docstring coverage — an AST proves it. Tags: docs, code-quality, agent-ops Run it: `curl -X POST https://app.loopskill.io/api/loops/doc-coverage-loop/run` (empty body OK; returns `passed: true/false`) - `json-schema-validate-loop` — JSON-Schema-Validate Loop: Drive a data file until it validates against its schema. Tags: data-validation, ci, agent-ops Run it: `curl -X POST https://app.loopskill.io/api/loops/json-schema-validate-loop/run` (empty body OK; returns `passed: true/false`) - `secret-scan-loop` — Secret-Scan Loop: Block the 'pushed an API key' incident: scan until clean. Tags: security, ci, agent-ops Run it: `curl -X POST https://app.loopskill.io/api/loops/secret-scan-loop/run` (empty body OK; returns `passed: true/false`) ## Composite loops — scheduled, multi-step (deploy once, runs nightly) A composite loop is a *standing* agentic routine: a composition of steps plus its own verifier plus a cadence. You do not POST-run these ad hoc — you place a composite loop onto an agent in your fleet and it runs on schedule from then on, verifying its own output each cycle. This is the tier that turns a skill catalog into an operating agent. - List composite loops (no key): `GET https://app.loopskill.io/api/composite-loops` - Detail (full composition): `GET https://app.loopskill.io/api/composite-loops/{slug}` - Deploy to a fleet agent (session required): `POST https://app.loopskill.io/api/composite-loops/{slug}/deploy` - `repo-stewardship-pack` — Repo Stewardship Pack: Install once, wake up to a repo that triages itself, reviews itself, and won't let the test suite go red. Tags: agent-ops, ci, github, code-quality, scheduled Cadence: every 30m · verified by `repo-steward-loop` Deploy it: https://app.loopskill.io/loops/view?slug=repo-stewardship-pack (or `POST https://app.loopskill.io/api/composite-loops/repo-stewardship-pack/deploy` with a signed-in session and `{fleet_id, member_id}`) - `atomic-habits` — Atomic Habits — 1% daily improvement loop: Ships one verified 1% improvement to your agent every night — compounding gains while you sleep, no babysitting. Tags: self-improvement, agent-ops, daily, compounding, scheduled Cadence: every 24h · verified by `test-green-loop` Deploy it: https://app.loopskill.io/loops/view?slug=atomic-habits (or `POST https://app.loopskill.io/api/composite-loops/atomic-habits/deploy` with a signed-in session and `{fleet_id, member_id}`) - `dreaming` — Dreaming — nightly memory consolidation: Your agent's nightly sleep cycle: consolidates the day's memories, extracts what mattered, prunes the noise — so tomorrow it starts sharper. Tags: self-improvement, memory, consolidation, agent-ops, scheduled Cadence: every 24h · verified by `test-green-loop` Deploy it: https://app.loopskill.io/loops/view?slug=dreaming (or `POST https://app.loopskill.io/api/composite-loops/dreaming/deploy` with a signed-in session and `{fleet_id, member_id}`) ## Bundles — curated skill collections (install one, get many) A bundle groups multiple skills (and connectors) into one install. Bundles are the "playlist" primitive — public bundles are browsable and installable with no key. - List public bundles (no key): `GET https://app.loopskill.io/api/bundles/discover` - Detail: `GET https://app.loopskill.io/api/bundles/public/{slug}` - `dev-agent-essentials` — Dev Agent Essentials (3 skills): The core skill set for an autonomous coding agent. Adversarial code review and PR draft automation — everything a dev agent needs to review… - `research-and-report` — Research & Report (3 skills): A client-ready report, on a schedule. Search the web, summarise what matters, and publish a formatted deliverable your agent hands off… - `loopskill-essentials` — LoopSkill Essentials (53 skills): The curated first-party catalog in one install: every skill published and quality-gated by the LoopSkill team. Agent ops, infra watchdogs,… - `coreys-marketing` — Corey's Marketing (50 skills): The complete marketing operating system for your agent — 40+ conversion, SEO, copywriting, paid, growth, and RevOps skills that work… - `agent-marketing` — agent_marketing (25 skills) - `seo-and-search-growth` — SEO & Search Growth (8 skills): Grow organic traffic without a marketing team. On-page audits, programmatic SEO, local SEO, and AI-search optimization — the 'seo' / 'seo… - `copywriting-and-humanizer` — Copywriting & Content Humanizer (8 skills): Write copy that converts, then make it sound human. The most repeated zero-result search on LoopSkill was 'copywriting' — this bundle is… - `cold-outreach-and-email` — Cold Outreach & Email Sequences (8 skills): Write the emails that get replies. Cold-email openers, multi-step sequences, and a composer agents can hand a full campaign to — the exact… - `ripwire` — Ripwire (18 skills): Stop making your agent read the whole repo. Ripwire's 17 skills wire an agent into a ranked symbol map, a blast radius, and a tests-to-run… - `terraform-and-kubernetes-ops` — Terraform & Kubernetes Infra Ops (8 skills): Provision and run infrastructure without leaving the agent loop. Terraform module generation and review paired with Kubernetes manifest and… - `agent-bench` — Agent Bench (0 skills): A loop-safe specialist bench for multi-agent fleets: Scout (recon, read-only), Builder (implements, never merges its own work), Reviewer… - `proactive-and-autonomous-agent-ops` — Proactive & Autonomous Agent Ops (8 skills): Agents that act before you ask, and improve on their own runs. The single loudest unserved-demand signal in LoopSkill's search log —… - `swarm-refactor-campaign` — Swarm Refactor Campaign (3 skills): Run a Teknium-style autonomous refactor swarm on your own repo — gate stack first (frozen-surface parity, identical-red-set baseline, AST… - `open-code-review` — Open Code Review (2 skills): Review the diff before you commit it. Alibaba ran this reviewer internally for two years across tens of thousands of engineers; it is a… ## Personalities — system-prompt archetypes for your agent A personality is a versioned system-prompt archetype (research analyst, focused dev agent, etc.) you can install onto an agent, distinct from a skill (a capability) or a loop (a runnable routine). 5 public personalities today. - List public personalities (no key): `GET https://app.loopskill.io/api/personalities` - Detail: `GET https://app.loopskill.io/api/personalities/{slug}` - `research-analyst` — Research Analyst [free]: A methodical research persona. Cross-references multiple sources, flags uncertainty, and cites everything. Great for due-diligence,… - `reviewer-independent-critic` — Reviewer — Independent Critic [free]: An independent code-review persona for multi-agent fleets. Produces reproducible verdicts with severity labels (blocking / should-fix /… - `scout-recon-specialist` — Scout — Recon Specialist [free]: A read-only reconnaissance persona for multi-agent fleets. Evaluates OSS, competitors, prior art and demand signals, then hands findings to… - `focused-dev-agent` — Focused Dev Agent [free]: A disciplined software engineer persona. Ships clean, tested code, asks clarifying questions before writing, and never hallucinates library… - `builder-implementation-specialist` — Builder — Implementation Specialist [free]: An implementation persona for multi-agent fleets. Turns a spec or handoff into working, tested code and owns implementation files, branches… ## Connectors — MCP-server config fragments A connector is a named, versioned MCP-server config template (stdio/http/sse) — literal secrets never transit the server, only ${VAR} env refs. Catalogued alongside skills and bundles. - List public connectors (no key): `GET https://app.loopskill.io/api/connectors` - Detail: `GET https://app.loopskill.io/api/connectors/{slug}` > Note: the public connector catalog is intentionally empty until a human promotes an entry — connectors are staged behind a review gate. See https://app.loopskill.io/docs/scope. ## Start free - [copywriting](https://app.loopskill.io/skills/copywriting): free — Write and edit marketing copy that survives scrutiny — landing pages, emails, ads, product descriptions, and launch posts. Direct-response… - [humanizer](https://app.loopskill.io/skills/humanizer): free — Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based… - [agentic-os](https://app.loopskill.io/skills/agentic-os): free — Operating constitution for autonomous AI-agent fleets: four laws, four seats, trust ledger (earned autonomy), goals-as-predicates, cost… - [Super Memory](https://app.loopskill.io/skills/super-memory): free — One-command installer for the full agent memory stack: cognee 1.0.x knowledge graph (pghybrid), LiteLLM proxy with rotation, cognee-api… - [ruthless-mentor](https://app.loopskill.io/skills/ruthless-mentor): free — Stress-test the team's plan, idea, or decision in attack mode — sort proposals into gold / trash / directionally-right-but-flawed and push… - [Coldstart Bench 20260908 B8C56156](https://app.loopskill.io/skills/coldstart-bench-20260908-b8c56156): free — Throwaway benchmark skill for the LoopSkill agent cold-start eval. Says hello. Safe to delete/ignore. - [elevenlabs-pro](https://app.loopskill.io/skills/elevenlabs-pro): free — Advanced ElevenLabs TTS — text-to-speech, voice listing/filtering, credits management. Script-based, works standalone. - [hundred-million-offers](https://app.loopskill.io/skills/hundred-million-offers): free — Create irresistible offers using the Value Equation, bonus stacking, risk-reversing guarantees, and ethical scarcity. Use when the user… - [critical-code-reviewer](https://app.loopskill.io/skills/critical-code-reviewer): free — Conduct rigorous, adversarial code reviews with zero tolerance for mediocrity. Use when users ask to 'critically review' my code or a PR,… - [ascii-video](https://app.loopskill.io/skills/ascii-video): free — ASCII video: convert video/audio to colored ASCII MP4/GIF. - [clean-code](https://app.loopskill.io/skills/clean-code): free — Write readable, maintainable code through disciplined naming, small functions, and clean error handling. Use when the user mentions 'code… - [musk-5-step-algorithm](https://app.loopskill.io/skills/musk-5-step-algorithm): free — The 5-step engineering algorithm (question requirements, delete, simplify, accelerate, automate — in that strict order) as an operating… - [xitter](https://app.loopskill.io/skills/xitter): free — Interact with X/Twitter via the x-cli terminal client using official X API credentials. Use for posting, reading timelines, searching… - [llama-cpp](https://app.loopskill.io/skills/llama-cpp): free — llama.cpp local GGUF inference + HF Hub model discovery. - [creative-ideation](https://app.loopskill.io/skills/creative-ideation): free — Generate project ideas via creative constraints. - [plan-for-goal](https://app.loopskill.io/skills/plan-for-goal): free — Author a goal-execution plan-doc that is ready to paste into your goal/issue tracker as N discrete, falsifiable goals. Use when the… - [llm-wiki-hermes](https://app.loopskill.io/skills/llm-wiki-hermes): free — Karpathy's LLM Wiki: build/query interlinked markdown KB. - [Hermes Swarm Refactor Campaign](https://app.loopskill.io/skills/hermes-swarm-refactor-campaign): free — Run a Teknium-style autonomous refactor swarm with a gate stack first: parity harness, red-set baseline, AST differentials, nested Hermes… - [brand-rollout-meta-repo](https://app.loopskill.io/skills/brand-rollout-meta-repo): free — Build and roll out a stack-agnostic brand meta-repo across multiple consumer products. Use when unifying design/brand/assets across… - [hub-search-claude-code](https://app.loopskill.io/skills/hub-search-claude-code): free — Scanner skill for LLM agents that need to discover locally-installed Claude Code skills and community npm plugins before installing a… - [loopskill](https://app.loopskill.io/skills/loopskill): free — Cross-vendor skill marketplace for AI agents (Hermes, Claude Code, Codex, OpenClaw, Cursor). Search, install, and run verified skills and… - [whisper](https://app.loopskill.io/skills/whisper): free — OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language… - [faster-whisper](https://app.loopskill.io/skills/faster-whisper): free — Local, offline speech-to-text using CTranslate2 reimplementation of OpenAI Whisper. 4-6x faster, 99+ languages, word-level timestamps, VAD… - [hyperspace-matrix](https://app.loopskill.io/skills/hyperspace-matrix): free — Search 100k+ indexed AI skills, tools, agents, and apps by natural language task description. 'Google for AI capabilities. ## Pricing Free: self-host the whole platform yourself (MPL-2.0, open source, no card needed). Pro — $9.95/mo: we host it for you (managed registry + runner, auto-updated catalog, ed25519-signed delivery). Pro is convenience only, never a feature gate — every capability that exists on Pro also exists in the free self-host. Founding Member — $49 one-time: permanent Pro, no subscription, never billed again. Everything in Pro, locked in at the founding price forever. Only 100 of 100 founding seats left. - Pricing page: https://app.loopskill.io/pricing ## Featured skills - [agentic-os](https://app.loopskill.io/skills/agentic-os) [free]: Operating constitution for autonomous AI-agent fleets: four laws, four seats, trust ledger (earned… - [ruthless-mentor](https://app.loopskill.io/skills/ruthless-mentor) [free]: Stress-test the team's plan, idea, or decision in attack mode — sort proposals into gold / trash /… - [Coldstart Bench 20260908 B8C56156](https://app.loopskill.io/skills/coldstart-bench-20260908-b8c56156) [free]: Throwaway benchmark skill for the LoopSkill agent cold-start eval. Says hello. Safe to… - [hundred-million-offers](https://app.loopskill.io/skills/hundred-million-offers) [free]: Create irresistible offers using the Value Equation, bonus stacking, risk-reversing guarantees, and… - [critical-code-reviewer](https://app.loopskill.io/skills/critical-code-reviewer) [free]: Conduct rigorous, adversarial code reviews with zero tolerance for mediocrity. Use when users ask… - [xitter](https://app.loopskill.io/skills/xitter) [free]: Interact with X/Twitter via the x-cli terminal client using official X API credentials. Use for… - [llama-cpp](https://app.loopskill.io/skills/llama-cpp) [free]: llama.cpp local GGUF inference + HF Hub model discovery. - [copywriting](https://app.loopskill.io/skills/copywriting) [free]: Write and edit marketing copy that survives scrutiny — landing pages, emails, ads, product… - [humanizer](https://app.loopskill.io/skills/humanizer) [free]: Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound… - [Super Memory](https://app.loopskill.io/skills/super-memory) [free]: One-command installer for the full agent memory stack: cognee 1.0.x knowledge graph (pghybrid),… ## Key pages - Catalog: https://app.loopskill.io/skills - Seeker (scan local installs, diff vs catalog): https://app.loopskill.io/seeker (tool: `loopskill_seeker`) - Skillify (runbook → installable skill): https://app.loopskill.io/skillify (tool: `loopskill_skillify`) - Share tokens (hand a client one install command): https://app.loopskill.io/docs/share-tokens - Runnable loops: https://app.loopskill.io/browse?type=loops (API: https://app.loopskill.io/api/loops) - Composite loops (scheduled): https://app.loopskill.io/browse?type=loops (API: https://app.loopskill.io/api/composite-loops) - Docs (install + MCP wiring): https://app.loopskill.io/docs - Pricing: https://app.loopskill.io/pricing - Compatibility (supported agents): https://app.loopskill.io/compatibility - Blog (architecture + product notes): https://app.loopskill.io/blog - Sitemap: https://app.loopskill.io/sitemap.xml ## About LoopSkill runs on a head-chef + line-cooks model: one orchestrating agent delegates to specialist skills. Skills are signed, versioned, and run with no cloud round-trip at execution time.