ginuim/skill-base / skills/skill-base-cli
skill-base-cli
UnverifiedThe official Skill Base CLI client. Use the `skb` (Skill Base CLI) command to search, install (single skill or collection), update, delete, publish, and import-from-GitHub skills from Skill Base, as well as configure skb and manage multiple environment profiles (e.g. prod / qa servers, each with its own login). Triggered when users say "publish skill to skill base", "install collection", "import skill from github", "download/update/delete skill from skill base", "configure skb", "add/switch skb profile", "publish to qa / prod skill base", "am I logged in", "check skb login", or "configure skill-base-cli".
Pull and Publish Skills with skb
Assistants should complete operations by running terminal commands. Package name skill-base-cli, command skb after installation.
When to Use This Skill
- Users request to search, install, update, delete, or publish specific skills using the
skbcommand. - Users want to install an admin-curated collection (recommended skill pack) in one shot.
- Users need to configure the client connection address (
skb init/skb config) or log in (skb login). - Users work with multiple Skill Base environments (e.g. prod and qa) and want to add, view, switch, or target a specific one.
- Users want to search, install, update, or publish Skills to a privately deployed Skill Base site.
When NOT to Use This Skill
- Users want to deploy, start, or maintain the Skill Base server itself (use
skill-base-web-deployinstead).
Web detail pages offer AI Agent prompts and CLI commands bound to the selected version and site. When using these, skip CLI installation if already available and choose the current agent's install directory via skb install --help.
Web contributor profiles at /users/:id list skills the user published versions for. Publishing via skb publish or GitHub import contributes to the same history; no separate CLI profile command is required.
The Web publish form shows the final included and excluded files before upload. CLI publishing supports preview and ignore rules; GitHub-import arguments are unchanged.
Environment
- Node.js >= 18
- Installation:
npm install -g skill-base-cli, ornpx skill-base-cli <subcommand> - Server address resolution:
SKB_BASE_URLenv var > active profile in~/.skill-base/config.json> defaulthttp://127.0.0.1:8000 - Active profile resolution: global flag
--profile <name>> env varSKB_PROFILE>currentProfilesaved byskb config use - Set and save the active profile's server:
skb init --server <site root URL>(do not include/api)
Multiple Environments (Profiles)
Each profile is a named server URL with its own login token, so logging in to qa does not overwrite prod. Tokens live in ~/.skill-base/credentials.json keyed by profile name.
skb config list # or: skb config ls; "*" marks the active profile
skb config show [name] # details of a profile (default: active); token is masked
skb config add qa -s https://qa.example.com [--use] # add; --use switches to it
skb config set qa -s <new_url> # change server URL
skb config set qa --rename test # rename (credentials move with it)
skb config use qa # switch current profile (alias: switch)
skb config remove qa [-y] # delete profile and its saved token (alias: rm)
list/showsupport--jsonfor scripts.- Run any command against another profile without switching:
skb --profile qa <command>(e.g.skb --profile qa login,skb --profile qa publish ./my-skill), orSKB_PROFILE=qa skb <command>. - An explicitly selected profile that does not exist makes the command fail and lists available profiles.
- When the user names an environment ("publish to qa", "install from prod"), check
skb config listfirst and pass--profile <name>rather than switching the user's current profile. - Profile names: letters, digits,
.,_,-. - Legacy single-server configs are read as a profile named
default.
Login and Authentication Rules (Important)
- No login required:
search,install,update,initand other regular read operations usually do not require login. Assistants should not proactively ask users to log in before these operations. - Login required:
skb publishandskb import-githubmust be logged in. - Login flow (
skb login):- Execute
skb loginin the terminal. - The console will output a login page URL containing
from=cli, open the login page in a browser. - After successful web login, a verification code (in the format
XXXX-XXXX) will be provided. - Enter the verification code in the terminal to exchange for a PAT.
- Execute
- Login, logout and whoami act on the active profile only; use
skb --profile <name> loginto log in to another environment. The verification code prompt is interactive, so the user must run it themselves. - Logout:
skb logout(clears only the active profile's token). - Check login:
skb whoami(callsGET /api/v1/auth/mewith the saved PAT; exit0if valid; output includes the profile). Useskb whoami -qfor scripts (exit code only) orskb whoami --jsonfor machine-readable output.skb config listshows login status of every profile.
Local Web UI
skb ui— starts a small server on127.0.0.1:9847(override with--port), opens the browser, proxies/api/v1/*to the configured Skill Base with the saved PAT. Use--no-open,--host <address>(default127.0.0.1).
Search and Install
skb search <keyword>
skb install <skill_id> # Latest version
skb install <skill_id>@<version> # Specific version, e.g., v20260327.161122
skb install <skill_id> -d <target_directory>
skb install <skill_id> -i cursor # Install to IDE skill directory
skb install <skill_id> -i cursor -g # Global IDE directory (where supported)
Options: -d/--dir, -i/--ide (alias -a/--agent), -g/--global. Common IDE values: cursor, claude-code, codex, copilot, windsurf, qoder, opencode, trae, openclaw, codebuddy, workbuddy, and others — run skb install --help for the full list.
Install a Collection
Collections are admin-curated flat packs (e.g. "frontend team essentials"), not a folder tree. One command downloads the collection zip and installs all member skills.
skb install --collection <id_or_slug>
skb install --collection 1 --ide cursor
skb install --collection frontend-team --ide cursor
target(skill id) and--collectionare mutually exclusive- On name conflicts, CLI lists all conflicting paths and asks once whether to overwrite all; cancel skips the entire collection install
Manage Local Installs
skb list # or: skb ls
skb update <skill_id>
skb update <skill_id> -d <directory>
skb delete <skill_id> # or: skb rm
skb delete <skill_id> --all
skb delete <skill_id> -d <directory> -y
skb installrecords install path, version, timestamp, IDE type, and global flag in~/.skill-base/config.jsonskb list/skb lsaggregate by skill (most recent first); pick one to update, delete local files, or clear records onlyskb update <skill_id>shows versions with changelog and uploader, then multi-select recorded install directories (or updates directly when only one exists)skb update <skill_id> -d <directory>bypasses records and updates<directory>/<skill_id>directlyskb delete <skill_id>lists recorded install directories for multi-select (or--all/-d/-yfor scripting); only removes paths whose directory name matchesskill_id
Publish
- Skill name must pass
/^[\w-]+$/validation, e.g., skill-base-cli - Prepare content in a folder named after the skill, root directory must contain
SKILL.md - Recommended: Use frontmatter in SKILL.md to mark name and description as the skill name and description
- If SKILL.md frontmatter has a name field, it must match the folder name
- If no frontmatter is used: Use the folder name as the skill name; the first paragraph of non-
#text under the title is the skill description (can be overridden with--description) - Execute in the skill directory:
skb publish; or execute from any location:skb publish <skill_folder_path> - Common usage:
skb publish <path> --changelog "description" - The Skill root's
.gitignoreand.skillbaseignoreapply by default. Use repeatable--ignore <pattern>for one run, or--no-gitignorewhen the user intentionally wants files ignored by Git..cursorand.codebuddyare included unless an ignore rule excludes them. - Run
skb publish <path> --dry-runwith the same ignore options first. It needs no login and lists included and excluded paths without uploading. - Before an agent-initiated real publish, show the final file list and target Skill ID to the user and ask for approval. An explicit user request that already identifies the target and contents supplies that approval; do not ask twice.
- After publishing, describe the result to the user in 2 sentences, no need to be verbose
Import from GitHub (public repos, skill-base login required)
skb import-github owner/repoorskb import https://github.com/owner/repo,likehttps://github.com/anthropics/skills/tree/main/skills/pdf- Same auth as publish (
skb login) - Options:
--ref,--subpath,--target <skill_id>,--changelog,--dry-run(preview JSON only) - Server downloads the repo archive; private repos are not supported
Troubleshooting Failures
- First confirm the active profile (
skb config show) points to the instance the user wants to use, and thatSKB_BASE_URL/SKB_PROFILEare not overriding it unexpectedly, as skill-base can be deployed internally and ports can be changed, verify the address is correct - If the user deployed skill-base on a server, most servers need inbound/outbound port configuration to be accessible, users need to check if the address is accessible in the browser
- Check if it's a network issue, can use ping/telnet to confirm
401/ "Authentication required": the token of that profile is missing or expired — re-runskb login(with--profile <name>for a non-active profile)- Publish failure: First
skb loginfor the target profile; then verify the profile server URL, folder name and frontmatternameare consistent and match[\w-]+, checkSKILL.mdand network - Install failure: Use
skb searchto verifyskill_id; when unsure of version, useskb install <skill_id>to install latest
Web-generated commands leave ordinary URLs and skill IDs unquoted; arguments with spaces or shell separators use double quotes for Windows CMD compatibility. If a server URL was saved with literal single quotes, rerun skb init --server "https://your-server". Collection detail pages also offer a direct-ZIP AI Agent prompt to install each skill separately.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| webapp-testinganthropics/skillsToolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs. | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| browser-testing-with-devtoolsaddyosmani/agent-skillsTests in real browsers via Chrome DevTools MCP. Use when building or debugging anything that runs in a browser. Use when you need to inspect the DOM, capture console errors, analyze network requests, profile performance, or verify visual output with real runtime data. Requires the chrome-devtools MCP server to be configured. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| ci-cd-and-automationaddyosmani/agent-skillsAutomates CI/CD pipeline setup. Use when setting up or modifying build and deployment pipelines. Use when you need to automate quality gates, configure test runners in CI, or establish deployment strategies. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |