A Claude Code plugin marketplace. Plugins cover the full development lifecycle: release automation, infrastructure verification, Home Assistant integration dev, Qt UI development, GitHub repo health, plugin testing, and three-layer documentation propagation.
Add this marketplace to your Claude Code installation:
/plugin marketplace add L3DigitalNet/Claude-Code-PluginsOr using the full URL:
/plugin marketplace add https://github.com/L3DigitalNet/Claude-Code-Plugins.gitThen install individual plugins:
/plugin install release-pipeline@l3digitalnet-pluginsAuto-update keeps plugins current automatically. To verify it's enabled:
- Run
/pluginin Claude Code - Go to the Marketplaces tab
- Select l3digitalnet-plugins
- Look for Disable auto-update (meaning it's already on)
When auto-update is enabled, Claude Code refreshes the marketplace catalog and updates installed plugins at the start of each session.
Manual update if you prefer to control when updates happen:
# Refresh the marketplace catalog
/plugin marketplace update l3digitalnet-plugins
# Then update individual plugins via /plugin → Installed tab → Update now| Plugin | Type | Command | Description |
|---|---|---|---|
| GitHub Repo Manager | Commands + Skills | /repo-manager |
Conversational GitHub repo health assessment and maintenance |
| Handoff | Skills | /handoff:save, /handoff:load |
Save and load task context across machines via shared network drive |
| Home Assistant Dev | Commands + Skills + MCP | varies | Full HA integration development toolkit with 27 skills |
| Nominal | Commands | /preflight, /postflight, /abort |
Infrastructure verification session contract: 11 systems check security, reachability, backups, monitoring, and more |
| Opus Context | Skills + Hooks | always-on | Teaches Opus 4.6 to use its full 1M context window instead of conservative small-model defaults |
| Plugin Test Harness | MCP | 18 tools | Iterative test/fix/reload loop for plugin development |
| qdev | Skills | /research, /quality-review, /deps-audit, /doc-sync, /spec-update |
Development quality toolkit: pre-build research sweeps, convergence-loop quality reviews, CVE dependency audits, and inline doc sync |
| Qt Suite | MCP + Commands + Skills + Agents | /qt-suite:scaffold, /qt-suite:coverage, /qt-suite:visual |
Complete Qt development and testing toolkit: proactive agents, 16 skills, scaffolding, and headless GUI testing |
| Release Pipeline | Commands + Skills | /release |
Semver releases with pre-flight checks and changelog generation |
| Repo Hygiene | Commands | /hygiene |
Autonomous maintenance sweep for .gitignore, manifests, and READMEs |
| Test Driver | Commands + Skills | /test-driver:analyze, /test-driver:status |
Proactive testing via gap analysis, convergence loops, and persistent status tracking |
| Up Docs | Skills + Agents | /up-docs:repo, /up-docs:wiki, /up-docs:notion, /up-docs:all, /up-docs:drift |
Update documentation across three layers via dispatched sub-agents (Haiku propagators + Sonnet drift auditor) from session context, plus full infrastructure drift analysis |
These principles apply across all plugins in this collection. Individual plugins may define additional principles scoped to their domain.
[P1] Act on Intent: Invoking a command is consent to its implied scope. When intent is ambiguous, clarify scope before executing, not after. When intent is clear, execute without narration or confirmation of the obvious. A confirmation gate is warranted only when an operation is both truly irreversible and its scope materially exceeds what the invocation implies, not for operations that are merely large or look dangerous. If scope materially changes mid-execution, stop and re-confirm. Routine friction is a tax, not a safeguard.
[P2] Scope Fidelity: Execute the full scope of what was asked, completely, without routine sub-task confirmation gates. When a sub-task's scope is genuinely ambiguous, clarify before proceeding rather than assuming. Do not act beyond the declared scope; surface only material out-of-scope findings as notes; routine adjacent observations are not worth raising. Scope undershoot triggers additional iteration; scope overshoot violates the consent established at invocation.
[P3] Succeed Quietly, Fail Transparently: Lead with findings, not intent or preamble. Output results, not narration. Emit a compact factual summary at task completion, not after every step. On transient or infrastructure errors, retry silently. On critical failures, stop immediately and surface the complete failure (raw output included) with a recovery plan; do not attempt autonomous workarounds.
[P4] Use the Full Toolkit: When interaction is required, prefer structured choices over open-ended prompts; bound the user's decision space before presenting it. Use Claude Code's interaction primitives: bounded choices (AskUserQuestion), markdown previews for side-by-side comparisons, multi-select for non-exclusive options.
[P5] Convergence is the Contract: Iterative work defines completion as a measurable criterion (set by the plugin, the user, or collaboratively) and drives toward it without check-ins. Proceed quietly when converging normally. Surface immediately if progress stalls or regresses unexpectedly. If the cycle begins oscillating (making and undoing the same changes repeatedly), flag the pattern and stop rather than continuing. Stop only when the criterion is met, oscillation is detected, or the user intervenes.
[P6] Composable, Focused Units: Every plugin component (command, skill, hook) does one thing and is independently useful. Complex workflows emerge from combining atomic units at runtime; orchestration is assembled from the outside, not baked in.
Conversational GitHub repository maintenance: assess and fix repo health interactively, with owner approval at every step.
Features:
- Wiki sync: keeps wiki pages in sync with in-repo docs and code
- Community health: audits and updates CONTRIBUTING, SECURITY, CODE_OF_CONDUCT, issue/PR templates
- PR triage: conflict detection, staleness checks, review summaries
- Issue triage: labels, assignees, staleness, linked PRs
- Release health: unreleased commits, changelog drift, draft releases
- Security posture: Dependabot alerts, code scanning, secret scanning
- Dependency audit: outdated packages, license concerns
- Notifications & discussions: triage and summarise
- Cross-repo mode: scan all accessible repos for a specific concern and fix in batch
- Structured maintenance report generated at session end
Install:
/plugin install github-repo-manager@l3digitalnet-pluginsLearn more: plugins/github-repo-manager/README.md
Cross-machine task continuity: save task context to a shared network drive when work requires physically moving between machines, then load it on the other side.
Features:
/handoff:save: writes a structured handoff file with task summary, current state, and next steps/handoff:load: reads the most recent handoff file and presents actionable next steps- Files stored on shared network path for access from any machine
Install:
/plugin install handoff@l3digitalnet-pluginsLearn more: plugins/handoff/README.md
Home Assistant integration development toolkit: 27 AI skills, an MCP server for live HA connections, automated validation, example integrations, and project templates.
Features:
- 27 context-aware skills covering architecture, config flows, coordinators, entities, testing, and more
- 3 specialized agents (development, review, debugging)
- MCP server with 12 tools for live HA connection and documentation search
- 5 validation scripts with PostToolUse hook enforcement
- 3 example integrations (Bronze/Silver/Gold tier)
- 9 project templates for CI/CD, testing, and documentation
- Full Integration Quality Scale coverage (all 52 rules)
Install:
/plugin install home-assistant-dev@l3digitalnet-pluginsLearn more: plugins/home-assistant-dev/README.md
Structured verification routine for infrastructure changes: three slash commands enforce a session contract ensuring every change begins with a validated environment and ends with a fully verified outcome.
Features:
/preflight: automated environment discovery (Mission Survey), go/no-go poll, rollback readiness confirmation/postflight: runs all 11 verification systems covering operational scripts, backup integrity, secrets hygiene, reachability, security posture, performance baselines, boot ordering, observability, DNS/certs, network routing, and documentation state/abort: confirmed rollback execution with step-by-step verification and post-abort go/no-go poll- Multi-environment support in a single profile
- Append-only flight log based on the OpenTelemetry Log Data Model
- Fix-forward flow with regression sweep to catch side effects
- Grounded in ITIL, CIS Controls v8, NIST SP 800-190, Google SRE PRR, and HashiCorp/OWASP secrets management
Install:
/plugin install nominal@l3digitalnet-pluginsLearn more: plugins/nominal/README.md
Iterative plugin testing framework: generates tests, records pass/fail results, applies source fixes, reloads the target plugin, and retests until convergence.
Features:
- Auto-generates tests from plugin source and schema introspection
- Test/fix/reload loop with convergence tracking (improving, plateau, oscillating, diverging)
- Dedicated git branch per session for a complete audit trail of fixes
- Sessions persist to disk and can be resumed after interruption
- Native MCP client for MCP-mode plugins; source analysis for hook/command plugins
- 18 tools across session management, test management, execution, and fix management
Install:
/plugin install plugin-test-harness@l3digitalnet-pluginsLearn more: plugins/plugin-test-harness/README.md
Autonomous release pipeline: quick merge or full semver release with parallel pre-flight checks, changelog generation, and GitHub release creation.
Features:
- Two modes: quick merge (testing → main) or full versioned release
- Parallel pre-flight agents (test runner, docs auditor, git validator)
- Automatic changelog generation from conventional commits
- Version bumping across Python, Node.js, Rust, and plugin manifests
- GitHub release creation with release notes
- Human-in-the-loop approval gates at critical stages
- Fail-fast with rollback guidance on errors
Install:
/plugin install release-pipeline@l3digitalnet-pluginsLearn more: plugins/release-pipeline/README.md
Autonomous maintenance sweep: five parallel mechanical checks plus a semantic README pass, with safe corrections applied automatically and destructive changes requiring explicit approval.
Features:
.gitignorepattern validation against actual repo contentsmarketplace.jsonpath and format checks- Orphaned plugin cache detection
- Stale
enabledPluginsentry cleanup - Semantic README freshness scan with inline AI reasoning
--dry-runflag shows full plan without touching anything
Install:
/plugin install repo-hygiene@l3digitalnet-pluginsLearn more: plugins/repo-hygiene/README.md
Complete Qt development and testing toolkit: proactive specialist agents, 16 domain skills, scaffolding commands, and headless GUI testing via the bundled Qt Pilot MCP server. Covers PySide6, PyQt6, and C++/Qt.
Features:
- 4 proactive agents: development specialist, debugger, code reviewer, and UX advisor
- 16 context-aware skills covering signals/slots, layouts, Model/View, threading, QML, styling, and more
/qt-suite:scaffold: generates a complete PySide6 project with pyproject.toml, src layout, and test config/qt-suite:generate: scans codebase and generates unit tests for untested files/qt-suite:coverage: gcov/lcov (C++) or coverage.py (Python) report with gap-targeted test generation/qt-suite:visual: launches app headlessly via Xvfb, drives UI via the bundled Qt Pilot MCP server
Install:
/plugin install qt-suite@l3digitalnet-pluginsLearn more: plugins/qt-suite/README.md
Development quality toolkit: five skills covering the full dev cycle from initial research through delivery.
Features:
/research: sweeps current libraries, APIs, and best practices before you build/quality-review: iterates to convergence, running checks until findings drop to zero/deps-audit: scans dependency manifests for known CVEs and outdated packages/doc-sync: aligns inline documentation with current implementation/spec-update: propagates code changes back to the specification document
Install:
/plugin install qdev@l3digitalnet-pluginsLearn more: plugins/qdev/README.md
1M context window optimizer for Opus 4.6: always-on behavioral rules that override conservative small-model defaults (partial reads, excessive delegation, re-reading).
Features:
- Whole-file reading by default (no offset/limit for files under 4000 lines)
- Direct reading over subagent delegation
- Dependency pre-loading before editing
- Budget-aware context pressure management
Install:
/plugin install opus-context@l3digitalnet-pluginsLearn more: plugins/opus-context/README.md
Proactive testing awareness and gap filling: always-on testing mindset that suggests gap analysis at natural breakpoints, finds missing tests across six categories, and iterates through a convergence loop until everything passes.
Features:
- Always-on testing-mindset skill triggers at natural breakpoints
- Six-category gap analysis (unit, integration, e2e, contract, security, UI)
- Convergence loop with oscillation detection
- Stack profiles for FastAPI, Django, PySide6, Home Assistant, and Swift
- Persistent test status tracking across sessions
Install:
/plugin install test-driver@l3digitalnet-pluginsLearn more: plugins/test-driver/README.md
Three-layer documentation updater via sub-agent dispatch: infers what changed during a session and updates repo docs, Outline wiki, and Notion at the right level of detail for each layer. Also provides comprehensive drift analysis that SSHes into live infrastructure.
Features:
- Parallel sub-agent architecture: three Haiku propagators (repo, wiki, notion) run in isolated context windows for cost efficiency (~1/10 Opus cost), while Sonnet audit ensures drift detection quality
/up-docs:repo,/up-docs:wiki,/up-docs:notion,/up-docs:all: dispatch targeted propagators from session context/up-docs:drift: four-phase convergence loop that gathers live server state via SSH, syncs the Outline wiki, resolves internal contradictions, verifies and enriches links, then updates Notion- Wall-clock time to completion reduced to
max(repo, wiki, notion)via parallel dispatch; sequential drift audit phases protect data integrity
Install:
/plugin install up-docs@l3digitalnet-pluginsLearn more: plugins/up-docs/README.md
The marketplace uses principle-traceable test coverage across all in-scope plugins. See testing/STRATEGY.md for the canonical frameworks per language (bats for bash, pytest for Python, Jest for TypeScript) and the principle → enforcement-layer mapping that guides test scope.
Quick reference:
# Bash plugins
cd plugins/release-pipeline && ./tests/run-bats.sh
# Python plugins
pytest plugins/home-assistant-dev/tests/ -m unit
# TypeScript plugins
cd plugins/plugin-test-harness && npm ci && npm run build && npm test
# Marketplace schema validation (always run before merging to main)
./scripts/validate-marketplace.shEach plugin's per-plugin plan lives in testing/plans/<plugin>.md with
coverage status and Phase 2 execution logs.
This repository also serves as a development workspace for creating new plugins. See CLAUDE.md for architectural guidance and docs/ for the full reference.
-
Create a new plugin:
mkdir -p plugins/my-plugin/.claude-plugin cd plugins/my-plugin -
Add manifest:
{ "name": "my-plugin", "version": "0.1.0", "description": "Plugin description" } -
Test locally:
claude --plugin-dir ./plugins/my-plugin
-
Add to marketplace catalog (
.claude-plugin/marketplace.json)
- docs/plugins.md - Plugin development guide
- docs/plugin-marketplaces.md - Marketplace creation
- docs/plugins-reference.md - Technical reference
- docs/skills.md - Creating AI-invoked skills
- docs/sub-agents.md - Custom agent definitions
- docs/hooks.md - Lifecycle event handlers
- docs/mcp.md - MCP server integration
Claude-Code-Plugins/
├── .claude-plugin/
│ └── marketplace.json # Marketplace catalog
├── plugins/ # All plugin implementations (12 plugins)
│ ├── github-repo-manager/ # Conversational GitHub repo maintenance
│ ├── handoff/ # Cross-machine task continuity (save/load)
│ ├── home-assistant-dev/ # Home Assistant integration dev toolkit
│ ├── nominal/ # Infrastructure verification (preflight/postflight/abort)
│ ├── opus-context/ # 1M context window optimizer for Opus 4.6
│ ├── plugin-test-harness/ # Iterative plugin testing framework
│ ├── qdev/ # Development quality toolkit (research, reviews, dep audits, doc-sync, spec-update)
│ ├── qt-suite/ # Qt development and testing toolkit (agents, skills, MCP)
│ ├── release-pipeline/ # Autonomous release pipeline
│ ├── repo-hygiene/ # Autonomous repo maintenance sweep
│ ├── test-driver/ # Proactive testing via gap analysis and convergence
│ └── up-docs/ # Three-layer documentation updater (repo, wiki, Notion)
├── scripts/
│ └── validate-marketplace.sh # Marketplace validation
├── docs/ # Comprehensive documentation
├── CLAUDE.md # Development guidance for AI agents
├── BRANCH_PROTECTION.md # Branch protection and workflow guide
└── README.md # This file
To add a plugin to this marketplace:
- Create plugin in
plugins/directory - Add entry to
.claude-plugin/marketplace.json(version must match the plugin's ownplugin.json) - Validate with
./scripts/validate-marketplace.sh - Commit directly to
mainand push - To publish a tagged release with GitHub release notes, run
/release-pipeline:releaseand pick the appropriate mode
Branch workflow: Direct commit to main. There is no testing branch. Local pre-commit hooks (noreply email enforcement, marketplace validation) provide guardrails. See BRANCH_PROTECTION.md for full rules.
git pull origin main
# (make edits)
./scripts/validate-marketplace.sh
git add <specific files>
git commit -m "..."
git push origin mainMIT - See LICENSE file for details
![@dependabot[bot]](https://avatars.githubusercontent.com/in/29110?s=64&v=4){kind=small}