{
  "markdown": "# Cekura AI Skills\n\nAI-powered skills for building and improving voice agent tests and metrics on the [Cekura](https://cekura.ai) platform. Works with Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, and other AI coding assistants.\n\n## Table of Contents\n\n- [What's Included](#whats-included)\n- [Prerequisites](#prerequisites)\n- [Quick Install (npx skills add)](#quick-install-npx-skills-add)\n- [Claude Code (VS Code)](#claude-code-vs-code)\n- [Claude Code (Terminal CLI)](#claude-code-terminal-cli)\n- [Codex](#codex)\n- [Cursor](#cursor)\n- [Gemini CLI](#gemini-cli)\n- [GitHub Copilot](#github-copilot)\n- [Windsurf / Other Agents](#windsurf--other-agents)\n- [MCP Server](#mcp-server)\n- [Quick Reference](#quick-reference)\n- [Platform Compatibility](#platform-compatibility)\n- [Reinstalling Cekura skills](#reinstalling-cekura-skills)\n- [Links](#links)\n\n---\n\n## What's Included\n\n### 12 Skills, 14 Commands in one plugin\n\n| Skills | Commands |\n|--------|----------|\n| `cekura-coordinator` | `cekura-onboarding`, `setup-mcp`, `upgrade-skills`, `report-bug` |\n| `cekura-onboarding` | `create-metric`, `list-metrics`, `evaluate-calls`, `improve-metric` |\n| `cekura-create-agent` | `manual-create-update-eval`, `autogen-eval`, `list-evals`, `run-evals`, `eval-results`, `cekura-report` |\n| `cekura-self-improving-agent` | \"Improve my agent / auto-tune from eval results\" — **also** \"fix a production call bug end-to-end\". Works on dashboard-managed providers AND custom stacks (config in your repo/DB/prompt registry, runtime-created provider agents, custom mock servers) via a per-project capability manifest |\n| `cekura-metric-design` | |\n| `cekura-metric-improvement` | |\n| `cekura-predefined-metrics` | |\n| `cekura-eval-design` | |\n| `cekura-infra-test-suite` | |\n| `cekura-flag-call-log-failures` | |\n| `cekura-generate-scenarios` | |\n\nThese encode best practices from real client deployments — proactive guardrails, real transcript grounding, iterative improvement loops, coverage planning, and anti-pattern detection.\n\n> **Stuck after an upgrade?** If `/upgrade-skills` reports stale plugin entries or `/plugin install` fails with `Source path does not exist`, do a clean reinstall — see [Reinstalling Cekura skills](#reinstalling-cekura-skills) below.\n\n## Prerequisites\n\n- **Cekura account** — [Sign up here](https://dashboard.cekura.ai/sign-up). Sign in via OAuth or use an API key.\n- **Cekura API key** *(only for the Claude Code plugin path / programmatic MCP access)* — Found under Settings > API Keys in the [Cekura dashboard](https://dashboard.cekura.ai). Not needed for `npx skills add`.\n\n---\n\n## Quick Install (`npx skills add`)\n\nThe fastest way to get Cekura skills into any [Agent Skills](https://agentskills.io)–compatible client (Claude Code, Cursor, Codex, Windsurf, OpenCode, and many more).\n\n### Install\n\n```bash\nnpx skills add cekura-ai/cekura-skills\n```\n\nThe CLI prompts you to pick which skills to install and which agents to install them into. To install everything non-interactively:\n\n```bash\nnpx skills add cekura-ai/cekura-skills --all\n```\n\n### Update\n\n```bash\n# Refresh existing skills\nnpx skills update\n\n# Or stay fully current — refresh existing AND pick up any newly-added skills\nnpx skills add cekura-ai/cekura-skills --all\n```\n\n### Remove\n\n```bash\nnpx skills remove cekura-coordinator   # one skill\nnpx skills remove --all                 # everything\n```\n\n> **GitHub Copilot:** Copilot loads skills from a repository's `.github/skills/` directory, so target it explicitly — `npx skills add cekura-ai/cekura-skills --all --agent github-copilot --output .github/skills`. For the Copilot CLI plugin (skills **and** MCP tools), see [GitHub Copilot](#github-copilot) below.\n\n### What gets installed\n\nTwelve skills, scoped to specific Cekura workflows:\n\n| Skill | When it activates |\n|---|---|\n| `cekura-coordinator` | \"What can Cekura do?\" — routes you to the right skill |\n| `cekura-onboarding` | \"Get started with Cekura\" — full platform walkthrough |\n| `cekura-create-agent` | \"Connect my voice agent to Cekura\" |\n| `cekura-self-improving-agent` | \"Improve my agent / auto-tune from eval results\" — **also** \"fix a production call bug / reproduce and test a fix before raising a PR\" |\n| `cekura-metric-design` | \"Create a metric / measure call quality\" |\n| `cekura-metric-improvement` | \"Improve a metric / fix metric accuracy\" |\n| `cekura-predefined-metrics` | \"What predefined metrics are available / which built-in metrics should I use\" |\n| `cekura-eval-design` | \"Design test scenarios for my voice agent\" |\n| `cekura-infra-test-suite` | \"Create a committed JSON CI/CD suite / add Tests-as-Code to my voice-agent repo / update test coverage for this PR\" |\n| `cekura-flag-call-log-failures` | \"Analyze the last N calls for issues / what % of calls have <problem>\" |\n| `cekura-generate-scenarios` | \"Create scenarios from failed calls / regression-test the agent on prod issues\" |\n\n### Want full functionality?\n\n`npx skills add` gives you the **behavioral guidance layer** — the skills auto-activate when you describe relevant tasks. For slash commands and direct API integration, install the full Claude Code plugin marketplace below.\n\n---\n\n## Claude Code (VS Code)\n\nFull plugin support — skills, slash commands, MCP tools, and auto-configured API access.\n\n### Install\n\n1. Open the Claude Code chat panel\n2. Click **Manage Plugins** > **Marketplaces** tab\n3. Paste `https://github.com/cekura-ai/cekura-skills.git` and click **Add**\n4. Switch to the **Plugins** tab > search for `cekura` > install the plugin\n5. In the Claude Code chat, run `/setup-mcp` to connect the MCP server. The default is OAuth — a browser opens for one-click sign-in; no API key or restart needed.\n6. (Optional) For the API-key auth path or curl-based skill fallbacks, set your key and restart VS Code:\n   ```bash\n   # Add to ~/.zshrc or ~/.bashrc\n   export CEKURA_API_KEY=\"your-key-here\"\n   ```\n\n### Get Started\n\nAsk \"I'm new to Cekura, help me get started\" for a guided walkthrough, or ask \"what can Cekura do?\" to see everything available.\n\n### Upgrade\n\nRun `/upgrade-skills` in any Claude Code session — it refreshes the marketplace and re-pins the plugin to the latest version, then prompts you to run `/reload-plugins` (no restart needed in the common case).\n\nTo upgrade manually:\n\n```bash\nclaude plugin marketplace update cekura-skills   # refresh catalog from GitHub\nclaude plugin update cekura@cekura-skills          # move the installed pin to latest\n```\n\nThen run `/reload-plugins` in your session to apply it. A plain `git pull` of the marketplace checkout does **not** move the version pin, so it won't upgrade you on its own — use the commands above.\n\n#### Auto-update\n\nInstalls from this marketplace (`cekura@cekura-skills`) keep themselves current: the plugin ships a `SessionStart` hook that runs Claude Code's own update commands at most once per day (best-effort — it never blocks or fails your session, and it no-ops for installs from any other marketplace). To also let Claude Code's native auto-update refresh the marketplace, add this to your `~/.claude/settings.json` — or just run `/setup-mcp`, which offers to enable it for you:\n\n```json\n{\n  \"extraKnownMarketplaces\": {\n    \"cekura-skills\": {\n      \"source\": { \"source\": \"github\", \"repo\": \"cekura-ai/cekura-skills\" },\n      \"autoUpdate\": true\n    }\n  }\n}\n```\n\n---\n\n## Claude Code (Terminal CLI)\n\nSame full plugin support as VS Code.\n\n### Install\n\n1. Inside a Claude Code session, run `/plugins`\n2. Go to the **Marketplaces** tab > select **Add Marketplace**\n3. Paste `https://github.com/cekura-ai/cekura-skills.git` and confirm\n4. Go to the **Discover** tab > search for `cekura` > install the plugin\n5. Run `/setup-mcp` to connect the MCP server. The default is OAuth — a browser opens for one-click sign-in; no API key or restart needed.\n6. (Optional) For the API-key auth path or curl-based skill fallbacks, set your key and restart your terminal + Claude Code session:\n   ```bash\n   # Add to ~/.zshrc or ~/.bashrc\n   export CEKURA_API_KEY=\"your-key-here\"\n   ```\n\n> **Tip:** If you've already cloned the repo locally, you can paste the local path instead of the GitHub URL.\n\n### Get Started\n\nAsk \"I'm new to Cekura, help me get started\" for a guided walkthrough, or ask \"what can Cekura do?\" to see everything available.\n\n### Upgrade\n\nRun `/upgrade-skills` in any Claude Code session — it refreshes the marketplace and re-pins the plugin to the latest version, then prompts you to run `/reload-plugins` (no restart needed in the common case).\n\nTo upgrade manually:\n\n```bash\nclaude plugin marketplace update cekura-skills   # refresh catalog from GitHub\nclaude plugin update cekura@cekura-skills          # move the installed pin to latest\n```\n\nThen run `/reload-plugins` to apply it. A plain `git pull` of the marketplace checkout does **not** move the version pin. To pull new versions automatically at launch, see [Auto-update (optional)](#auto-update-optional) under Claude Code (VS Code), or run `/setup-mcp`.\n\n---\n\n## Codex\n\nNative plugin support — skills **and** MCP tools. (Codex plugins don't carry slash commands; the skills cover those workflows — they activate by context or when you invoke one with `@`.)\n\n### Install (recommended)\n\nThree steps — add the marketplace, install the plugin, then authenticate the MCP:\n\n```bash\n# 1. Add the marketplace\ncodex plugin marketplace add cekura-ai/cekura-skills\n\n# 2. Install the plugin (cekura plugin from the cekura marketplace)\ncodex plugin add cekura@cekura\n\n# 3. Authenticate the Cekura MCP — opens a browser for OAuth sign-in (no API key)\ncodex mcp login cekura\n```\n\nPrefer the TUI? Run `codex`, open `/plugins` to install **cekura**, then `/mcp` to authenticate.\n\nVerify it worked:\n\n```bash\ncodex plugin list   # cekura should be listed as installed\ncodex mcp list      # cekura should show as connected after login\n```\n\n> If you added the marketplace before a recent release, refresh it first so the latest manifest is picked up: `codex plugin marketplace upgrade cekura`.\n>\n> **API-key alternative:** the bundled MCP config is OAuth-only. To use a key instead, add the server to `~/.codex/config.toml` with `http_headers = { \"X-CEKURA-API-KEY\" = \"your-key\" }` (or `env_http_headers` to read it from an env var).\n\n### Fallbacks (no MCP)\n\n**Skills only** — install the skill files directly:\n\n```bash\npython3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \\\n  --repo cekura-ai/cekura-skills \\\n  --path cekura/skills/cekura-onboarding \\\n         cekura/skills/cekura-create-agent \\\n         cekura/skills/cekura-self-improving-agent \\\n         cekura/skills/cekura-metric-design \\\n         cekura/skills/cekura-metric-improvement \\\n         cekura/skills/cekura-predefined-metrics \\\n         cekura/skills/cekura-eval-design \\\n         cekura/skills/cekura-infra-test-suite\n```\n\n**Behavior preset** — single-file domain knowledge (metric design, eval design, API reference, anti-patterns):\n\n```bash\ncurl -o AGENTS.md https://raw.githubusercontent.com/cekura-ai/cekura-skills/main/codex/AGENTS.md\n```\n\n### Get Started\n\nAsk Codex to help with Cekura metrics or evals — skills load automatically when the conversation matches, or invoke one explicitly with `@`.\n\n### Upgrade\n\n```bash\ncodex plugin marketplace upgrade cekura   # refresh the marketplace snapshot from GitHub\ncodex plugin add cekura@cekura            # re-pin to the new version (required — not optional)\n```\n\nBoth steps are needed: `marketplace upgrade` only refreshes the snapshot, and `codex plugin add` re-pins the installed plugin to it. Running just the first leaves you on the old version. Confirm with `codex plugin list` — the `cekura@cekura` row should show the new version.\n\n**Auto-update (optional).** The plugin ships a `SessionStart` hook that runs the two commands above automatically on each launch (throttled to once/day), so you stay current without typing them. Codex requires you to trust the hook once: run `codex`, open `/hooks`, and trust the Cekura **SessionStart** hook. After that it's automatic; the manual commands above remain available to force an update immediately.\n\nFor the fallbacks, re-run the skill installer / re-download `AGENTS.md`.\n\n---\n\n## Cursor\n\nNative plugin support — skills **and** MCP tools. (Cursor plugins don't carry Claude-style slash commands; the skills cover those workflows.)\n\n### Install (recommended)\n\nIn Cursor, go to **Settings > Plugins > Team Marketplaces > Add Marketplace > Import from Repo**, point it at `https://github.com/cekura-ai/cekura-skills`, then install **cekura**. Authenticate the Cekura MCP via OAuth when prompted.\n\n### Fallback (no MCP)\n\nUse the behavior preset as a rules file:\n\n```bash\ncurl -o .cursor/rules/cekura.md https://raw.githubusercontent.com/cekura-ai/cekura-skills/main/codex/AGENTS.md\n```\n\nOr add it as a global rule in Cursor Settings > Rules.\n\n### Get Started\n\nAsk Cursor to help with Cekura metrics or evals — skills load automatically when the conversation matches.\n\n### Upgrade\n\n**Manual:** re-import the marketplace from **Settings > Plugins > Team Marketplaces** (or re-download the rules-file fallback).\n\n**Auto Refresh (optional)** — have Cursor pull new versions automatically:\n\n1. Install the **Cursor GitHub App** on the `cekura-ai/cekura-skills` repository (Cursor prompts for this; it's required for Auto Refresh).\n2. Go to **Settings > Plugins > Team Marketplaces** and locate the imported Cekura marketplace.\n3. Toggle on **Auto Refresh** and save.\n\nCursor then re-indexes the marketplace at most once every ~10 minutes, picking up pushed updates to the installed plugin. Note: Auto Refresh updates the *existing* plugin only — if a brand-new plugin is added to the repo later, re-import the marketplace to pick it up.\n\n**Team-wide auto-install (Teams/Enterprise):** a workspace admin can mark the Cekura plugin **Required** for a distribution group in the Team Marketplace settings — that installs and keeps it updated for everyone automatically, no per-developer action. This is an admin dashboard setting, not something the plugin declares.\n\n---\n\n## Gemini CLI\n\nExtension support — MCP tools plus the Cekura context file. (Gemini discovers extension skills from a root `skills/` directory, so this extension doesn't bundle the nested `cekura/skills/`; native Gemini skill bundling is deferred.)\n\n### Install\n\n```bash\ngemini extensions install https://github.com/cekura-ai/cekura-skills --auto-update\n```\n\nThe `--auto-update` flag lets the extension self-check GitHub on each launch and apply the new version on the next restart — no manual updates needed. Omit it if you prefer to update manually.\n\nOn first use of a Cekura MCP tool, Gemini runs an OAuth sign-in flow. The extension loads `GEMINI.md` — all Cekura domain knowledge (metric design, eval design, API reference, anti-patterns) — as context.\n\n### Upgrade\n\nIf you installed with `--auto-update`, new versions are pulled automatically (restart to apply). To update manually, or to force it now:\n\n```bash\ngemini extensions update cekura\n```\n\n---\n\n## GitHub Copilot\n\nCekura installs on Copilot two different ways, and they cover different surfaces:\n\n| Surface | Install | What you get |\n|---|---|---|\n| **Copilot CLI** | Marketplace plugin (below) | Skills **+** MCP tools, OAuth |\n| **Coding agent, code review, IDE extensions** | Skills committed to `.github/skills/` | Skills; MCP is configured per repository |\n\nCopilot reads its own registry at `.github/plugin/marketplace.json`, separate from the Claude manifest, so the two installs never interfere. Copilot plugins don't carry Claude-style slash commands — the skills cover those workflows.\n\n### Install (Copilot CLI)\n\n**1. Install the CLI and sign in** (needs a Copilot licence — Pro, Pro+, Business, or Enterprise):\n\n```bash\nnpm install -g @github/copilot\n```\n\nRun `copilot` once and complete the sign-in prompt.\n\n**2. Add the Cekura marketplace:**\n\n```bash\ncopilot plugin marketplace add cekura-ai/cekura-skills\n```\n\n**3. Confirm Copilot can read it** — this should list one plugin, `cekura`:\n\n```bash\ncopilot plugin marketplace browse cekura-skills\n```\n\n**4. Install the plugin:**\n\n```bash\ncopilot plugin install cekura@cekura-skills\n```\n\n**5. Verify skills and MCP:**\n\n```bash\ncopilot plugin list   # cekura listed as installed\n```\n\n```bash\ncopilot mcp list      # cekura listed as an MCP server\n```\n\n**6. Authenticate MCP** — nothing to configure. The first time Copilot calls a Cekura tool it opens a browser for OAuth sign-in; no API key is stored.\n\nPrefer an interactive session? Run `copilot`, then `/plugin marketplace add cekura-ai/cekura-skills` followed by `/plugin install cekura@cekura-skills`.\n\n### Copilot coding agent, code review, and the IDE extensions\n\nThose surfaces load skills from a repository's `.github/skills/` directory rather than from CLI plugins. Install the skill files into the repo you want them active in and commit the result:\n\n```bash\nnpx skills add cekura-ai/cekura-skills --all --agent github-copilot --output .github/skills\n```\n\nTo give those surfaces Cekura MCP access as well, register `https://api.cekura.ai/mcp` in your repository's Copilot MCP configuration — see GitHub's docs on [configuring MCP servers for your repository](https://docs.github.com/en/copilot/how-tos/copilot-on-github/customize-copilot/configure-mcp-servers). Note that the Copilot coding agent and Copilot code review don't support OAuth-authenticated remote MCP servers, so use the API-key credential there: send your key in an `X-CEKURA-API-KEY` header. (Copilot CLI does support OAuth — that path needs no key.)\n\n### Get Started\n\nAsk Copilot to help with Cekura metrics or evals — skills load automatically when the conversation matches. A good first prompt:\n\n```plaintext\nUse the Cekura skills and MCP. List my Cekura agents, pick one, and propose 3 evaluators I should create first. Do not create anything until I approve.\n```\n\nIf Copilot answers in generic terms without naming your agents, the skills or the MCP server aren't loaded — re-check steps 3–5.\n\n### Upgrade\n\n```bash\ncopilot plugin update cekura\n```\n\nIf a release adds a brand-new skill, refresh the marketplace snapshot first: `copilot plugin marketplace add cekura-ai/cekura-skills` (re-adding refreshes it), then `copilot plugin update cekura`. For the `.github/skills/` path, re-run the `npx skills add` command above. Copilot has no auto-update hook yet, so run the update command when you want the latest version.\n\n---\n\n## Windsurf / Other Agents\n\nCopy `codex/AGENTS.md` to wherever your agent reads context files from (project root, `.windsurf/rules/`, etc.):\n\n### Install\n\n```bash\ncurl -o AGENTS.md https://raw.githubusercontent.com/cekura-ai/cekura-skills/main/codex/AGENTS.md\n```\n\nThe file contains all Cekura domain knowledge in a single portable format that works with any agent.\n\n### Upgrade\n\nRe-download:\n\n```bash\ncurl -o AGENTS.md https://raw.githubusercontent.com/cekura-ai/cekura-skills/main/codex/AGENTS.md\n```\n\n---\n\n## MCP Server\n\nAll plugins connect to the Cekura API through an MCP (Model Context Protocol) server. This gives structured access to 84+ Cekura API operations as typed tools.\n\n**For Claude Code users:** the plugin auto-configures the MCP server; on first tool use you'll get a one-click browser OAuth sign-in (no API key stored). Run `/setup-mcp` if tools aren't available — it verifies connectivity and, if you prefer a key-based credential (e.g. for CI), walks you through the `X-CEKURA-API-KEY` setup.\n\n**For Codex, Cursor, Gemini CLI, and GitHub Copilot:** the MCP server is wired up natively through each platform's plugin/extension manifest, authenticating via OAuth (no API key stored). The sign-in step differs per platform — in **Codex** run `codex mcp login cekura`; **Cursor** prompts for OAuth when you connect the server; **Gemini** and **Copilot CLI** run the OAuth flow on first tool use. For agents using only the `AGENTS.md` behavior preset (Windsurf, etc.), the MCP server is optional — the preset includes API reference with curl examples as a fallback.\n\n**How it works:** Claude Code, Codex, and Copilot CLI read the bundled `cekura/.mcp.json`; Cursor and Gemini declare the endpoint inline in their own manifests. All five point at the Cekura MCP server at `https://api.cekura.ai/mcp` (CI asserts they stay in sync). By default it authenticates via OAuth — on first use the client opens a browser for a one-click sign-in, with no API key stored. To use an API key instead, run `/setup-mcp` and choose the API-key path. See the [MCP overview](https://docs.cekura.ai/mcp/overview).\n\n---\n\n## Quick Reference\n\n### Slash Commands (Claude Code plugin only)\n\n| Command | What it Does |\n|---------|-------------|\n| `/cekura-onboarding` | Guided end-to-end setup — preflight, state-aware resume, walks through agent + metrics + first eval run |\n| `/setup-mcp` | Configure MCP server (run once after install) |\n| `/upgrade-skills` | Re-pin the plugin to the latest version, then `/reload-plugins` |\n| `/report-bug` | Report a bug — files GitHub issue, optionally attempts a fix |\n| `/create-metric` | Create or update a metric |\n| `/list-metrics` | List metrics for an agent or project |\n| `/evaluate-calls` | Run metrics on specific calls |\n| `/improve-metric` | Improve metric accuracy: feedback, labs, auto-improve |\n| `/autogen-eval` | Auto-generate evaluators (or bulk create from CSV/JSON) |\n| `/manual-create-update-eval` | Create or update a single evaluator with full field walkthrough |\n| `/list-evals` | List evaluators for an agent or project |\n| `/run-evals` | Execute test scenarios |\n| `/eval-results` | Check results from a test run |\n| `/cekura-report` | Full end-to-end quality report — generates 10 evals, runs them, produces structured analysis |\n\n### Skills (load automatically — both install paths)\n\n| Skill | When it activates |\n|-------|-------------------|\n| `cekura-coordinator` | \"What can Cekura do?\" — routes to the right skill |\n| `cekura-onboarding` | First-time setup, end-to-end platform walkthrough |\n| `cekura-create-agent` | Setting up an agent — provider, mock tools, KB, dynamic vars |\n| `cekura-self-improving-agent` | Auto-tuning an agent prompt from eval results — **and** fixing a production call bug end-to-end (auto-build reproduction harness, must-fail-first gate, fix, stochastic verify, regression sweep, PR / summary) |\n| `cekura-metric-design` | Designing or creating metrics |\n| `cekura-metric-improvement` | Improving an existing metric via feedback iteration |\n| `cekura-predefined-metrics` | Exploring built-in metrics — what each does, costs, constraints |\n| `cekura-eval-design` | Designing test scenarios for a voice agent |\n| `cekura-infra-test-suite` | Creating or updating source-controlled Cekura JSON Tests-as-Code suites for a voice-agent repository |\n| `cekura-flag-call-log-failures` | Triaging recent production call logs against KPIs — failure rates + outcome distribution |\n| `cekura-generate-scenarios` | Turning flagged production failures into regression evaluator scenarios |\n\n### Getting Started Flow\n\n1. `/setup-mcp` — Configure API access (Claude Code plugin only)\n2. `/cekura-onboarding` — Guided platform setup (preflight + state-aware walkthrough). Or just ask \"I'm new to Cekura\" to activate the skill directly.\n3. Ask \"set up my agent\" — activates `cekura-create-agent`\n4. `/autogen-eval` — Auto-generate test scenarios\n5. `/run-evals` — Run your first tests\n6. Ask \"create a metric for X\" — activates `cekura-metric-design`\n7. `/cekura-report` — Full end-to-end quality report for any agent\n\n---\n\n## Platform Compatibility\n\n| Platform | Recommended method | Skills | MCP Tools | Slash Commands |\n|----------|--------|--------|-----------|---------------|\n| **Any Agent Skills client** | `npx skills add` | Yes | No | No |\n| **Claude Code (VS Code)** | Marketplace install | Yes | Yes | Yes |\n| **Claude Code (CLI)** | `/plugins` install | Yes | Yes | Yes |\n| **Codex** | `codex plugin marketplace add` | Yes | Yes | No |\n| **Cursor** | Plugin marketplace install | Yes | Yes | No |\n| **Gemini CLI** | `gemini extensions install` | Context file only | Yes | No |\n| **GitHub Copilot (CLI)** | `copilot plugin marketplace add` | Yes | Yes | No |\n| **GitHub Copilot (coding agent / code review / IDE)** | `npx skills add --output .github/skills` | Yes | Repo MCP config | No |\n| **Windsurf** | Rules file | Behavior preset | No | No |\n| **Other agents** | Copy AGENTS.md | Behavior preset | No | No |\n\n---\n\n## Reinstalling Cekura skills\n\nIf `/upgrade-skills` reports stale plugin entries, or `/plugin install` fails with `Source path does not exist`, your Claude Code install needs a clean reinstall. Run these 4 commands in order in any Claude Code session:\n\n1. `/plugin marketplace remove cekura-skills`\n2. `/plugin marketplace add cekura-ai/cekura-skills`\n3. `/plugin marketplace update cekura-skills`\n4. `/plugin install cekura@cekura-skills`\n\nStep 3 is the critical one — it forces Claude Code to re-read the marketplace manifest. After step 4, `claude plugin list` should show one `cekura@cekura-skills` entry, and all 14 slash commands resolve under `cekura:*`.\n\n### Troubleshooting\n\n**`/plugin install` errors with `Source path does not exist`:** Re-run step 3 (`/plugin marketplace update cekura-skills`) and retry step 4. If that doesn't clear it, fully restart Claude Code (close and reopen the session), then redo steps 2–4.\n\n**`claude plugin list` shows zero cekura entries after upgrading:** you've removed the old plugins but haven't installed the new one. Run steps 2–4 above.\n\n---\n\n## Data & privacy\n\nThe plugin talks only to your own Cekura account via the Cekura MCP server (`https://api.cekura.ai/mcp`), with every call mediated by your client's MCP permission system.\n\n- **Skill-usage ping:** when a Cekura skill or command activates, it calls the `cekura_skill_started` MCP tool with the skill name, its verification tag, the plugin version, and (for some commands, when available) a conversation/session ID — so Cekura can see which skills are used and validate that the right playbook was loaded. Nothing else from your conversation is sent.\n- **Local failure log:** failures of Cekura MCP tools are logged to `~/.claude/cekura-mcp-failures.log` (secret-redacted, capped at 100 lines, never leaves your machine on its own). The `/report-bug` command may include redacted excerpts in a GitHub issue — only after showing you the full issue body and getting your explicit OK.\n- **Auto-update:** see the Auto-update sections above (Claude Code: daily self-update hook for installs from this marketplace; Codex: daily hook after a one-time `/hooks` trust).\n\nSee Cekura's [Privacy Policy](https://www.cekura.ai/privacy-policy) and [Terms of Service](https://www.cekura.ai/terms-of-service). Questions or issues: **support@cekura.ai**.\n\n---\n\n## Links\n\n- **Cekura Dashboard:** [dashboard.cekura.ai](https://dashboard.cekura.ai)\n- **Sign Up:** [dashboard.cekura.ai/sign-up](https://dashboard.cekura.ai/sign-up)\n- **API Docs:** [docs.cekura.ai/api-reference](https://docs.cekura.ai/api-reference)\n- **LLM-friendly Docs:** [docs.cekura.ai/llms.txt](https://docs.cekura.ai/llms.txt)\n- **Concepts:** [docs.cekura.ai/documentation/key-concepts](https://docs.cekura.ai/documentation/key-concepts/)\n",
  "bytes": 27279,
  "sha": "9cc21e396452ab0b96b72d75a4e912e7bc192d47767b67f260dcc638205d4bd9",
  "repo_slug": "cekura-ai/cekura-skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_cekura_ai_cekura_skills_962e5d23/readme"
}