Project documentation for Claude Code and AI assistants working on this repository.
Developer Tools Directory is a meta-repository that catalogs, standardizes, and scaffolds all TMHSDigital Cursor IDE plugins and MCP servers. It does not contain a plugin or MCP server itself.
License: CC-BY-NC-ND-4.0 outbound, DCO + broader inbound grant (see CONTRIBUTING.md)
Author: TMHSDigital
Catalog site: https://tmhsdigital.github.io/Developer-Tools-Directory/
Developer-Tools-Directory/
registry.json # Source of truth for all tool repos
standards/ # 18 convention docs (structure, CI/CD, testing, MCP, security, licensing, etc.)
scaffold/
create-tool.py # Python repo generator (Jinja2)
templates/ # Jinja2 templates for new repos
scripts/
sync_from_registry.py # Regenerates README, CLAUDE.md, and docs embedded registry
site-template/
build_site.py # Builds GitHub Pages for tool repos from their metadata
template.html.j2 # Shared HTML template with configurable branding
fonts/ # Self-hosted Inter + JetBrains Mono woff2
SETUP-PROMPT.md # Copy-paste prompt for applying template to a repo
docs/
index.html # GitHub Pages catalog site
style.css # Dark theme, responsive, card layout
script.js # Fetches registry.json, renders cards, filters
.well-known/security.txt # RFC 9116 security contact
contributors/ # Worked examples for new contributors
assets/
logo.png # Directory logo
.github/
CODEOWNERS # Review routing
PULL_REQUEST_TEMPLATE.md
ISSUE_TEMPLATE/ # Bug report, feature request, config
workflows/ # Validate, sync, pages, release, stale, codeql, etc.
README.md # Action-pinning convention
Array of tool objects. Each entry has: name, repo, slug, description, type (cursor-plugin | mcp-server), homepage, skills, rules, mcpTools, extras, topics, status, version, language, license, pagesType, hasCI. status is one of experimental, beta, active, maintenance, deprecated, or archived (see standards/lifecycle.md).
When updating, run python scripts/sync_from_registry.py to regenerate every derived artifact. The sync-check CI job blocks PRs that drift.
Pure-stdlib Python script that regenerates README tables and stats, CLAUDE.md cataloged tools and totals, and the embedded registry JSON in docs/index.html. Modes: no flag (write), --check (CI drift detection), --about (print the gh repo edit command for the GitHub About section, run locally).
CLI args: --name, --slug, --description, --type, --mcp-server, --skills N, --rules N, --license, --output, --author-name, --author-email.
Templates in scaffold/templates/ produce: plugin.json, GitHub Actions workflows, README.md, AGENTS.md, CLAUDE.md, CONTRIBUTING.md, CHANGELOG.md, CODE_OF_CONDUCT.md, SECURITY.md, ROADMAP.md, LICENSE, .cursorrules, .gitignore, docs/index.html, and optional mcp-server/ scaffold.
Static HTML/CSS/JS. No build step. No external CDN. The pages.yml workflow copies registry.json and assets/ into docs/ at deploy time. script.js fetches registry.json at runtime and falls back to an embedded copy. Never use innerHTML/eval with registry data - the safety-scan CI job blocks reintroduction.
17 Markdown documents defining conventions for every tool repo. Index in standards/README.md.
| Tool | Type | Skills | Rules | MCP Tools |
|---|---|---|---|---|
| CFX Developer Tools | Plugin | 9 | 6 | 6 |
| Unity Developer Tools | Plugin | 18 | 8 | 4 |
| Docker Developer Tools | Plugin | 17 | 10 | 150 |
| Home Lab Developer Tools | Plugin | 22 | 11 | 50 |
| Mobile App Developer Tools | Plugin | 43 | 12 | 36 |
| Plaid Developer Tools | Plugin | 17 | 7 | 30 |
| Monday Cursor Plugin | Plugin | 21 | 8 | 45 |
| Steam Cursor Plugin | Plugin | 30 | 9 | 25 |
| Steam MCP Server | MCP Server | 0 | 0 | 25 |
| Developer Tools MCP | MCP Server | 0 | 0 | 7 |
| CFX MCP | MCP Server | 0 | 0 | 0 |
Totals: 177 skills, 71 rules, 378 MCP tools across 10 repos
main is protected by the main protection ruleset. Direct pushes, force pushes, and branch deletion are all blocked for every contributor including the repo owner (empty bypass list). All changes land via pull request with squash merge only.
Standard loop for any change, including single-file edits:
git checkout -b <type>/<short-description>
# edits
git add <paths>
git commit -s -m "<conventional subject>"
git push -u origin HEAD
gh pr create --base main
gh pr checks <number> --watch
gh pr merge <number> --squash --delete-branchA PR cannot be merged until all 7 required status checks pass: Validate registry.json, Validate docs site, Validate scaffold, Registry sync check, Public-repo safety scan, feat/fix commits require VERSION bump, CodeQL. Required approvals are 0 (solo maintainer). release.yml's Check VERSION vs latest tag is a post-merge guard, not a PR gate.
pip install -r requirements.txt # Jinja2python scaffold/create-tool.py \
--name "Test Plugin" \
--description "Automated test" \
--mcp-server \
--skills 2 \
--rules 1 \
--output /tmp/testOpen docs/index.html in a browser. The script falls back to embedded registry data when not served from GitHub Pages.
validate.yml runs on every PR and push to main:
- Registry JSON syntax and schema validation
- Docs file existence checks
- Scaffold Python syntax check
- Scaffold dry-run test (generates repo, verifies key files exist, scans for leaked business email)
sync-checkjob: runspython scripts/sync_from_registry.py --checkand fails on driftsafety-scanjob: blocks leaked business emails, drive-letter paths, unsafe DOM sinks (innerHTML/evalindocs/), and committed credential filesversion-bump-checkjob (PR only): requires aVERSIONbump when the PR contains anyfeat:orfix:commit; opt out per commit with[skip version]
release.yml runs on push to main and reads the VERSION file at the repo root. If VERSION is ahead of the latest git tag, it creates tag v<VERSION> and a GitHub Release. If VERSION equals the latest tag, the workflow no-ops. If VERSION is lower, the workflow fails. Release notes come from release-drafter's current draft body when present, otherwise from the commit log since the previous tag. The meta-repo uses VERSION-file-driven releases; tool repos use conventional-commit auto-bumps per standards/versioning.md.
- Conventional commits with DCO sign-off (
git commit -s) - No em dashes, en dashes, or emoji - see
standards/writing-style.md - No hardcoded credentials, business emails, or local filesystem paths anywhere
- Public readership - all docs written for external consumption
- Single branch -
mainonly - Self-contained docs site - no CDN, no unsafe DOM sinks
registry.jsonis the single source of truth - README, CLAUDE.md, anddocs/index.htmlembedded data are generated
| Dependency | Purpose | Where |
|---|---|---|
| Jinja2 >=3.1,<4.0 | Scaffold template rendering | requirements.txt |
The docs site has zero runtime dependencies. The sync script is pure stdlib.
You have context-mode MCP tools available. These rules are NOT optional — they protect your context window from flooding. A single unrouted command can dump 56 KB into context and waste the entire session.
Any Bash command containing curl or wget is intercepted and replaced with an error message. Do NOT retry.
Instead use:
ctx_fetch_and_index(url, source)to fetch and index web pagesctx_execute(language: "javascript", code: "const r = await fetch(...)")to run HTTP calls in sandbox
Any Bash command containing fetch('http, requests.get(, requests.post(, http.get(, or http.request( is intercepted and replaced with an error message. Do NOT retry with Bash.
Instead use:
ctx_execute(language, code)to run HTTP calls in sandbox — only stdout enters context
WebFetch calls are denied entirely. The URL is extracted and you are told to use ctx_fetch_and_index instead.
Instead use:
ctx_fetch_and_index(url, source)thenctx_search(queries)to query the indexed content
Bash is ONLY for: git, mkdir, rm, mv, cd, ls, npm install, pip install, and other short-output commands.
For everything else, use:
ctx_batch_execute(commands, queries)— run multiple commands + search in ONE callctx_execute(language: "shell", code: "...")— run in sandbox, only stdout enters context
If you are reading a file to Edit it → Read is correct (Edit needs content in context).
If you are reading to analyze, explore, or summarize → use ctx_execute_file(path, language, code) instead. Only your printed summary enters context. The raw file content stays in the sandbox.
Grep results can flood context. Use ctx_execute(language: "shell", code: "grep ...") to run searches in sandbox. Only your printed summary enters context.
- GATHER:
ctx_batch_execute(commands, queries)— Primary tool. Runs all commands, auto-indexes output, returns search results. ONE call replaces 30+ individual calls. - FOLLOW-UP:
ctx_search(queries: ["q1", "q2", ...])— Query indexed content. Pass ALL questions as array in ONE call. - PROCESSING:
ctx_execute(language, code)|ctx_execute_file(path, language, code)— Sandbox execution. Only stdout enters context. - WEB:
ctx_fetch_and_index(url, source)thenctx_search(queries)— Fetch, chunk, index, query. Raw HTML never enters context. - INDEX:
ctx_index(content, source)— Store content in FTS5 knowledge base for later search.
When spawning subagents (Agent/Task tool), the routing block is automatically injected into their prompt. Bash-type subagents are upgraded to general-purpose so they have access to MCP tools. You do NOT need to manually instruct subagents about context-mode.
- Keep responses under 500 words.
- Write artifacts (code, configs, PRDs) to FILES — never return them as inline text. Return only: file path + 1-line description.
- When indexing content, use descriptive source labels so others can
ctx_search(source: "label")later.
| Command | Action |
|---|---|
ctx stats |
Call the ctx_stats MCP tool and display the full output verbatim |
ctx doctor |
Call the ctx_doctor MCP tool, run the returned shell command, display as checklist |
ctx upgrade |
Call the ctx_upgrade MCP tool, run the returned shell command, display as checklist |