Install and route through OpenSpace, the skill management layer for AI agents, so a host agent can retrieve/rank/load the right SKILL.md out of this jeo-skills catalog (~150 installed skills), then evaluate skill quality from real execution evidence and evolve skills via FIX/DERIVED/CAPTURED updates. Covers install-as-skill-finder, retrieve-a-skill, evaluate-quality, evolve-skills, and local-first hub share/import. Triggers on: openspace, skill finder, skill retrieval, find the right skill, rank skills, skill discovery, skill quality, evolve skill, FIX DERIVED CAPTURED, skill hub, openspace-mcp, DiscoverSkills, skill-discovery, delegate-task.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
The command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
File Explorer
8 files
Showing SKILL.md
SKILL.md
Source instructions ยท Read-only preview
name
openspace
description
Install and route through OpenSpace, the skill management layer for AI agents, so a host agent can retrieve/rank/load the right SKILL.md out of this jeo-skills catalog (~150 installed skills), then evaluate skill quality from real execution evidence and evolve skills via FIX/DERIVED/CAPTURED updates. Covers install-as-skill-finder, retrieve-a-skill, evaluate-quality, evolve-skills, and local-first hub share/import. Triggers on: openspace, skill finder, skill retrieval, find the right skill, rank skills, skill discovery, skill quality, evolve skill, FIX DERIVED CAPTURED, skill hub, openspace-mcp, DiscoverSkills, skill-discovery, delegate-task.
allowed-tools
Bash Read Write Edit Glob Grep
compatibility
Requires Python 3.12+ and an MCP-capable host (Claude Code, Codex, Cursor, OpenClaw, nanobot). Local search, evaluation, and evolution work without cloud access; sharing to the cloud hub needs an OpenSpace cloud agent key.
Use this skill when the real question is "how does this host agent find, trust, and
improve the right SKILL.md out of a large catalog?" โ not when the user just wants a
single skill executed.
OpenSpace's own README frames it as "The Skill Management Layer for AI Agents": your
skills keep growing, and OpenSpace helps you retrieve, evaluate, and evolve with
every run. In this repo, that job is specifically: act as the skill-finder /
skill-retrieval layer for jeo-skills โ a host agent installs OpenSpace once, then asks
it to search, rank, and load the right SKILL.md out of the ~150 skills already
installed under .agent-skills/ (and copied into $HOME/.agents/skills), instead of an
agent guessing from a flat skill list or a human maintaining routing rules by hand.
When to use this skill
The user wants to install OpenSpace so a host agent can act as a skill-finder over an existing skill catalog
The user (or an agent) needs to retrieve/rank the best-fit skill for a task out of many installed skills
The user wants to know which installed skills actually work based on real execution outcomes, not just descriptions
The user wants controlled skill evolution (FIX / DERIVED / CAPTURED) driven by evidence from real runs
The user wants to share or import skills through a local-first hub (package browsing, explicit import, trust-gated upload)
The user is choosing between OpenSpace and a plain catalog grep, a long-lived knowledge wiki, or per-repo agent memory
When not to use this skill
The user only needs plain keyword/grep lookup across the existing skill catalog โ route to codebase-search or semble; do not stand up a full MCP skill-management layer for a one-off lookup
The user wants long-lived, synthesized markdown knowledge (concepts, indexes, research notes) โ route to llm-wiki
The user wants per-repo agent memory (handoff notes, decision logs, manifests) that is not about skill selection at all โ route to opencontext
The user just wants to run one specific, already-known skill โ load that skill's SKILL.md directly; OpenSpace is for finding and judging skills, not a replacement for using them
Instructions
Step 1: Classify the job into one routing mode
Pick exactly one primary mode before doing anything else:
install-as-skill-finder โ wire OpenSpace into a host agent so it can search this catalog
retrieve-a-skill โ given a task, find/rank/load the right SKILL.md
evaluate-skill-quality โ judge whether an installed skill is trustworthy from execution records
git clone --filter=blob:none --sparse https://github.com/HKUDS/OpenSpace.git ~/.openspace/OpenSpace
cd ~/.openspace/OpenSpace
git sparse-checkout set --no-cone '/*''!/assets/'# skips the ~50 MB assets/ folder# A dedicated venv, not the system Python: `pip install -e .` against a Homebrew or# distro Python fails with PEP 668 "externally-managed-environment", which is how# openspace-mcp ends up missing on a machine that "installed successfully".
python3 -m venv ~/.agents/venvs/openspace # or: uv venv --python 3.12 ~/.agents/venvs/openspace
~/.agents/venvs/openspace/bin/python -m pip install -e .
~/.agents/venvs/openspace/bin/openspace-mcp --help# verify installationln -sf ~/.agents/venvs/openspace/bin/openspace-mcp ~/.local/bin/openspace-mcp
Requires Python 3.12+.
Or run the bundled installer, which performs the same steps non-interactively, copies the
two host skills (Step 4), and registers the MCP server (Step 3):
Step 3: Wire the MCP server, with OPENSPACE_HOST_SKILL_DIRS pointed at this repo's skill root
Register the openspace MCP server in every AI runtime installed on the machine โ
Claude Code and its Anthropic-compatible forks (kimi, glm/zai, deepseek, grok, qwen),
Codex, Gemini CLI, Cursor, OpenCode, and the pi / gjc / jeopi agent runtimes. For
jeo-skills, OPENSPACE_HOST_SKILL_DIRS must point at $HOME/.agents/skills โ that is
where this repo's installers copy skills for host agents to load, so that is the
directory OpenSpace should scan and rank against.
Use the bundled registrar instead of hand-editing each config:
bash scripts/register-openspace-mcp.sh --dry-run # preview every runtime it would touch
bash scripts/register-openspace-mcp.sh # merge in place
bash scripts/register-openspace-mcp.sh --force # overwrite an existing openspace entry
It writes mcpServers JSON (~/.claude.json, ~/.claude/claude_desktop_config.json,
~/.gemini/settings.json, ~/.qwen/settings.json, ~/.cursor/mcp.json,
~/.kimi/mcp.json, ~/.glm/mcp.json, ~/.zai/mcp.json, ~/.deepseek/mcp.json,
~/.pi/agent/mcp.json, ~/.gjc/agent/mcp.json, ~/.jeopi/agent/mcp.json), TOML
[mcp_servers.openspace] (~/.codex/config.toml, ~/.grok/config.toml), and OpenCode's
mcp block with type: local (~/.config/opencode/opencode.json). Existing files keep
their mode and are replaced atomically; symlinks and non-regular configs are refused; a
runtime whose config directory does not exist is skipped rather than invented. The
absolute venv binary path is written, so registration does not depend on ~/.local/bin
being on the agent's PATH (resolution order: $OPENSPACE_VENV/bin/openspace-mcp,
~/.local/bin/openspace-mcp, PATH, legacy ~/.openspace/venv/bin/openspace-mcp).
The entry it writes is equivalent to:
{"mcpServers":{"openspace":{"command":"$HOME/.agents/venvs/openspace/bin/openspace-mcp","toolTimeout":600,"env":{"OPENSPACE_HOST_SKILL_DIRS":"$HOME/.agents/skills","OPENSPACE_WORKSPACE":"$HOME/.openspace/OpenSpace","OPENSPACE_CLOUD_MODE":"local","OPENSPACE_CLOUD_API_KEY":"sk-xxx (optional, for cloud)"}}}}
OpenSpace supports three launch modes; prefer stdio for local use:
stdio โ keep command: "openspace-mcp" in the host config (simplest option)
Do not report installation success until openspace-mcp --help works and the MCP
client can see OpenSpace's tools; long execute_task calls need a toolTimeout of at
least 600 seconds.
Step 4: Copy the two host skills that teach the agent when/how to call OpenSpace
skill-discovery/SKILL.md teaches the host agent to search/discover skills;
delegate-task/SKILL.md teaches it to execute, fix, and upload. For jeo-skills, copy
both into $HOME/.agents/skills/ (this is exactly what scripts/install-openspace.sh
automates). No additional prompting is needed after that โ the host agent now knows
when to reach for OpenSpace instead of guessing.
Step 5: Retrieve/discover the right skill for a task
Once installed, ask the host agent (via the openspace MCP tools, or skill-discovery)
to search and rank skills for the task at hand instead of scanning the catalog by hand.
OpenSpace discovers skills from OPENSPACE_HOST_SKILL_DIRS, configured
skills.skill_dirs, project roots such as .openspace/skills, user roots such as
~/.openspace/skills, and finally its own bundled openspace/skills โ in that
precedence order. Every discovered skill passes check_skill_safety before it can be
loaded; skills with dangerous patterns (prompt injection, credential exfiltration) are
blocked and logged.
Both ranking stages are optional at runtime and fail silently. BM25 needs the
rank_bm25 package in the OpenSpace venv; the semantic re-rank needs an OpenAI-compatible
key (OPENAI_API_KEY / OPENROUTER_API_KEY, or a nanobot/OpenClaw host config). With
neither, search_skills still answers, every result carries score: 0.0, and the order is
arbitrary โ the skill finder is effectively a name lookup. Always verify with an intent
query ("scrape a javascript rendered page" should surface scrapling), never a name
query, which passes even when ranking is dead.
Step 6: Evaluate skill quality from real execution evidence
Do not trust a skill because its description reads well. OpenSpace's quality layer
tracks whether a skill was selected, applied, completed, or fell back, and whether
its underlying tools became unreliable, slow, or risky โ using actual task behavior as
evidence instead of self-reported claims.
Step 7: Evolve skills only when evidence demands it
OpenSpace evolves skills through three controlled triggers:
FIX โ repair a broken or outdated skill
DERIVED โ create a better or more specialized version from an existing skill
CAPTURED โ save one reusable subworkflow, but only when the source trace shows
both its execution and a separate validation of the claimed postcondition
New/evolved skills are provisional by default; independent successful use promotes
a skill to trusted, while an attributable failure demotes it. enabled controls
reuse independently from that trust lifecycle. Blocked or uncertain proposals stay as
audit-only candidates โ recurrence never auto-promotes them.
Step 8: Share or import skills through the local-first hub
Skills run and are evolved locally; the cloud is for discovery and review, not
execution:
openspace-download-skill # download a skill from the cloud
openspace-upload-skill --skill-dir /path/to/skill/dir # upload a trusted skill
Cloud upload requires the matching local SkillStore record to be trusted โ both
public and private uploads fail closed for provisional or unknown records. The local
trust state itself is never sent to the cloud. Optional cloud bootstrap:
Register the openspace MCP server in every installed runtime with bash scripts/register-openspace-mcp.sh (OPENSPACE_HOST_SKILL_DIRS=$HOME/.agents/skills, OPENSPACE_WORKSPACE=$HOME/.openspace/OpenSpace)
Copy skill-discovery/ and delegate-task/ into $HOME/.agents/skills/
Verify tools are visible and a lightweight local skill search works
Example 2: Find the right skill for a task
Input
I need to scrape a JS-rendered page โ which of our installed skills should I use?
Output sketch
Mode: retrieve-a-skill
Ask OpenSpace to search/rank skills for "JS-rendered page scraping" across OPENSPACE_HOST_SKILL_DIRS
Expect it to surface scrapling (or an equivalent) ranked above generic candidates, using BM25 + embedding hybrid ranking, then load that skill's SKILL.md
Example 3: Judge whether a skill is still reliable
Input
We've used the deployment-automation skill a dozen times. Is it actually working?
