Skip to content

About

Four open-source skills for local search, editor synonyms, explicit learning, and authored memory in Cursor and coding agents.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Terraphim Skills logo

Terraphim Skills Introduction

Bring local search, editor synonyms, explicit learning, and authored memory into your agent workflow. This repository packages four Apache-2.0 instruction skills and documents direct or custom routes for Cursor, Kimi Code, AutoClaw, ZCode, Hermes Agent, skills.sh, and Claude Code. A Grok Build marketplace submission is under review.

v0.2.7 release details · Choose an agent · Try the CLI workflow · Review permissions · 简体中文

Start here

  1. Install the separate commands with the dependency guide, or use the Homebrew commands below. The skills do not install executables for you.
  2. Add the instruction skills to your agent using a direct or custom route in the host table. A curated marketplace listing has its own approval process.
  3. Try a bounded search, then explicitly capture an authored lesson with the checked example. Review the written memory before using it elsewhere.

The v0.2.7 release contains four skills and a Cursor integration archive under a signed tag. Its macOS ARM64 native qualification and independent review pass for the exact release source; productionSignoff is PENDING_USER_SIGNOFF. Automatic failure capture is unavailable. This release and review do not establish Cursor Marketplace admission or qualify other native platforms.

The four skills

Skill Command family Purpose
terraphim-grep terraphim-grep Search bounded local code and documentation, offline by default.
terraphim-agent-learn terraphim-agent learn Inspect failures and capture verified operational corrections.
terraphim-agent-memory terraphim-agent memory Save authored lessons, inspect saved content and source tags, and retrieve memory.
terraphim-editor Editor UI → terraphim-grep Create and swap alternatives, export synonyms and search related decisions.

These instruction skills do not bundle CLI executables, secrets, or proprietary skill content. They remain useful without a marketplace account. The Cursor setup guide provides a read-first command and optional integration examples. Installing the instruction plugin does not activate MCP, hooks, rewriting, or capture; each optional action requires its own reviewed setup and matching qualification evidence.

Install the command dependencies

On macOS or Linux with Homebrew:

brew tap terraphim/terraphim
brew install terraphim-grep terraphim-agent
terraphim-grep --version
terraphim-agent --version
terraphim-agent learn --help
terraphim-agent memory --help

Homebrew versions can differ from the release used in an example. The linked graph search and authored-memory workflow was checked with Terraphim Clients 1.21.20; the guide links to the release archives and checksums. Verify the installed versions before using it. See the non-technical installation guide or the repository's dependency installation for updates, removal, non-Homebrew installation, signature checks, and PATH troubleshooting. The skills never install dependencies automatically.

Install in your agent

The routes below describe installation methods, not v0.2.7 host acceptance. The guide's historical v0.2.4 examples cover a three-skill pack; they do not qualify the current four-skill release on non-Cursor hosts. The macOS ARM64 Cursor native review passes, with production sign-off still pending.

Host Direct or custom route Curated status
Cursor Copy the repository locally. Prior submission recorded; public listing unverified.
Kimi Code Install the repository or add marketplaces/kimi.json as a custom marketplace. Direct route documented; no curated claim.
AutoClaw Import one checksummed ZIP per skill from the release through Skills & Connectors → Skills → Create → Add skill file. Local import documented; external catalog review is separate.
ZCode Add marketplaces/zcode.json as a personal marketplace and install terraphim-skills-intro. Curated listing awaits maintainer review.
Hermes Agent Add terraphim/terraphim-cursor-plugin as a tap, then inspect and install the four skills. Tap route documented; no curated claim.
skills.sh Run npx skills add terraphim/terraphim-cursor-plugin; DISABLE_TELEMETRY=1 opts out of anonymous telemetry. Direct repository route documented.
Claude Code Add this repository as a marketplace, then install the plugin as shown below. Custom marketplace route documented.
Kilo Integration is being prepared. Catalog admission and actual Kilo installation unqualified.
Grok Build The submission pins a signed release; see the distribution guide. Awaiting maintainer review.

Claude Code commands:

claude plugin marketplace add terraphim/terraphim-cursor-plugin
claude plugin install terraphim-skills-intro@terraphim-skills

Availability in a host's curated public marketplace is separate from direct or custom installation. See the distribution guide for exact install, update, discovery, and removal steps and current publication status.

Local Cursor installation

For local development, copy this repository into:

~/.cursor/plugins/local/terraphim-skills-intro

Cursor 3.20 rejects a symlink when its resolved target is outside ~/.cursor/plugins/local, so use a real directory:

cp -R /path/to/terraphim-cursor-plugin \
  ~/.cursor/plugins/local/terraphim-skills-intro

Reload Cursor, open Customize, and confirm that all four skills appear.

Use both command-line tools

Search with an editor-exported synonym graph, then save and read an authored lesson. The example includes a real exported Risk/Concern graph, a clean public corpus, a baseline search and saved JSON outputs:

terraphim-grep Concern --search-only --haystack all \
  --paths ./corpus --thesaurus ./risk-concern.knowledge-graph.json --json -n 3

terraphim-agent --format json memory capture \
  --content 'Risk requires a Decision: approve a second supplier as the mitigation.' \
  --provenance-tag 'public-project-brief' > captured-memory.json
memory_id="$(python3 -c 'import json; print(json.load(open("captured-memory.json"))["memory_id"])')"
terraphim-agent memory show "$memory_id" --json

Create the corpus and copy the graph using the linked guide before running these commands. The memory capture writes to the current user's OS-local store, independent of the current working directory; exporting a search graph does not select a memory role or storage scope.

Read the synonym-search walkthrough, agent memory walkthrough and learning/hooks walkthrough for the complete flows.

Example requests

  • “Use Terraphim Grep to find the authorization callback and show two lines of context.”
  • “Search project learnings for an earlier failed deployment command.”
  • “Save this authored lesson in local memory, then show its saved ID and content.”

The skills default to local, read-only operations. Persistent learning or memory changes remain explicit and scoped. Review permissions and side effects before enabling writes or configured model synthesis.

Continue with Terraphim Skills

The skills can support an engineering workflow from research through validation. The diagram shows that method; it is not an automatic feature of this plugin.

V-model showing research, design, specification, implementation, verification, and validation phases

Browse the Community, Core, and Premium catalogue. The public introduction contains no proprietary instructions or entitlement token. Following a related-skill link never creates a checkout or transaction. Access and purchase decisions remain on Terraphim's authenticated site and MCP services.

For guided setup, see the non-technical installation guide. For agent integration, see the MCP guide. Read the privacy policy and terms of service. The engineering approach is explained in the canonical Disciplined Engineering for AI Systems article.

Development

python3 scripts/generate_distribution.py --check
python3 -m unittest discover -s tests -p 'test_*.py'
python3 scripts/generate_distribution.py --autoclaw-dist dist/autoclaw

Licence

Apache-2.0. See LICENSE.

About

Four open-source skills for local search, editor synonyms, explicit learning, and authored memory in Cursor and coding agents.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages