Install Olostep skill folders directly into your AI coding agents from the command line.
- Skills land in a canonical store (
~/.agents/skills) and are symlinked or copied into each agent’s skills directory
- Tracks all installs in a lockfile at
~/.agents/.skill-lock.json
- Works with Cursor, Claude, Codex, Windsurf, Continue, and more
This feature is available via the Olostep CLI.
The skills
olostep add skills installs all 13. Each is a SKILL.md your agent reads to know what Olostep can do and when to use it.
Skills pair with the MCP server: the MCP server gives the agent the live tools, the skills give it the know-how for when and how to use them.
Setup
Start here — teaches the agent how to configure the Olostep MCP server.
| Skill | What it does |
|---|
setup | Configure the Olostep MCP server so all other skills work |
Core web data
The main Olostep capabilities — use these to give your agent live web access.
| Skill | What it does |
|---|
scrape | Turn one URL into clean markdown / HTML / JSON / text |
search | Live web search — results, answers, and in-site URL discovery |
answers | Cited, structured answers from live web data |
crawl | Autonomously crawl a whole site |
map | Discover every URL on a site |
batch | Scrape up to 10,000 URLs in parallel |
extract-schema | Scrape a page into structured JSON matching a schema |
Build & integrate
For agents helping developers add Olostep to a codebase or keep up with API changes.
| Skill | What it does |
|---|
integrate | Auto-install the Olostep SDK into a project |
docs-to-code | Scrape API docs and write working code from them |
migrate-code | Read a migration guide and update local code |
Research & debug
For agents doing research, making decisions, or debugging code errors.
| Skill | What it does |
|---|
research | Cited, comparative web research for a decision |
debug-error | Look up an error message against live GitHub / SO / docs |
Categories
Every skill belongs to one of three categories, which map to the --category flag:
usage — use Olostep’s features: scrape, search, answers, crawl, map, batch, extract-schema.
build — install and integrate Olostep into a codebase: setup (configure the MCP server) and integrate (add the Olostep SDK to a project).
workflow — produce a deliverable with Olostep: research, docs-to-code, migrate-code, debug-error.
Install just one category:
olostep add skills --category usage # core web-data skills
olostep add skills --category build # setup + integration skills
olostep add skills --category workflow # research + deliverable skills
Hosted copies
Each installed skill also references its hosted copy at https://www.olostep.com/skills/<name>/SKILL.md — an agent with web access can fetch the latest version (which may include capabilities added since install) and falls back to the local copy otherwise. Browse all hosted skills at olostep.com/skills.
Install
npm install -g olostep-cli
# Install all 13 skills into every detected agent
olostep add skills
# Or use the skills subcommand (alias)
olostep skills install
Or without installing the CLI:
npx -y olostep-cli@latest add skills
Common usage
# Install all skills
olostep skills install
# Refresh / update to latest
olostep skills update
# See what's installed and where
olostep skills list
# Remove all skills
olostep skills uninstall
Filter what gets installed:
# Only core web-data skills
olostep add skills --category usage
# Only setup + integration skills
olostep add skills --category build
# Cherry-pick
olostep add skills --skill scrape --skill search --skill setup
# Specific agents only
olostep add skills --agent cursor --agent claude
# Machine-readable output
olostep add skills --json
Options reference
| Option | Default | Description |
|---|
--login | — | Run browser login before installing |
--source <path> | CLI bundled skills | Skills source directory |
--cli-local-dir <path> | CLI/skills | Directory where source skills are synced for CLI-local use |
--agent <name> | — | Target a specific agent — repeatable |
--all-agents / --no-all-agents | --all-agents | Target all detected agents |
--global / --no-global | --global | Install into global agent skill dirs |
--canonical-dir <path> | ~/.agents/skills | Canonical storage location |
--agent-skills-dir <path> | — | Custom target dir (requires --no-global) |
--skill <name> | — | Include only this skill — repeatable |
--exclude <name> | — | Exclude this skill — repeatable |
--category | — | usage, build, or workflow |
--overwrite / --no-overwrite | --overwrite | Replace existing installs |
--link-mode <auto|symlink|copy> | auto | auto tries symlink first, falls back to copy |
--json | — | Machine-readable JSON output |
Validation rules:
--link-mode must be auto, symlink, or copy
--agent-skills-dir requires --no-global
--no-global requires --agent-skills-dir
- Unknown agent names will error
- An empty skill selection after
--skill / --exclude will error
JSON output shape
When --json is passed, the output looks like:
{
"sync": {
"plugin_source_dir": "/path/to/CLI/skills",
"cli_local_dir": "/path/to/CLI/skills"
},
"selected_skills": ["scrape", "search"],
"canonical_dir": "~/.agents/skills",
"lockfile_path": "~/.agents/.skill-lock.json",
"installed": [
{
"skill": "scrape",
"canonical_path": "~/.agents/skills/olostep-scrape",
"targets": [
{ "agent": "cursor", "mode": "symlink", "path": "~/.cursor/skills/olostep-scrape" },
{ "agent": "claude", "mode": "symlink", "path": "~/.claude/skills/olostep-scrape" }
]
}
]
}
Supported agents
| Agent | Key |
|---|
| Cursor | cursor |
| Claude | claude |
| Codex | codex |
| Windsurf | windsurf |
| Continue | continue |
| Augment | augment |
| Roo | roo |
| Gemini | gemini |
| Copilot | copilot |
| Factory | factory |
Use --all-agents (the default) to target all detected agents, or --agent <key> for specific ones.
olostep list skills
See which skills are installed and into which agents — without digging through files.
olostep list skills # human-readable summary
olostep list skills --json # machine-readable
olostep remove skills
Removes Olostep-installed skill folders from the canonical store and agent skill directories, and cleans up the lockfile.
# Remove all Olostep skills from all agents
olostep remove skills
# Remove a specific skill
olostep remove skills --skill research
# Remove from a specific agent only
olostep remove skills --agent cursor
# Machine-readable JSON output
olostep remove skills --json
| Option | Default | Description |
|---|
--agent <name> | — | Remove only from specified agent(s) — repeatable |
--all-agents / --no-all-agents | --all-agents | Target all detected agents |
--canonical-dir <path> | ~/.agents/skills | Canonical skills directory to remove from |
--agent-skills-dir <path> | — | Custom target skills directory for removal |
--skill <name> | — | Remove only matching skill(s) — repeatable |
--json | — | Machine-readable JSON output |
Safety: Only folders with the olostep- prefix are touched. Non-Olostep skill folders in agent directories are never modified. Lockfile cleanup only removes olostep--prefixed keys.
Skills are discovered by scanning subdirectories for a SKILL.md file. The file must include a YAML frontmatter block with name and description fields. Duplicate sanitized names are rejected.
---
name: my-skill
description: What this skill does and when to use it
---
# My Skill
...skill instructions here...
Naming convention
Installed folders always use the olostep- prefix — e.g. research becomes olostep-research. Names are sanitized to lowercase with invalid characters replaced by -. This prefix keeps Olostep-managed skills clearly identifiable and prevents conflicts with other tools.