- Fix marketplace.json: change source from relative path to GitHub URL
(was "skills/unraid", now "https://github.com/jmagar/unraid-mcp")
This resolves the "Invalid input" schema validation error when adding
the marketplace to Claude Code
- Refactor subscriptions autostart to use anyio.Path for async file checks
(replaces blocking pathlib.Path.exists() with async anyio.Path.exists())
- Update dependencies: anyio 4.11.0→4.12.1, attrs 25.3.0→25.4.0
- Remove set -e from validate-marketplace.sh to prevent early exit on
check failures, allowing the summary to always be displayed (PRRT_kwDOO6Hdxs5uvKrc)
- Fix marketplace.json source path to point to skills/unraid instead of
./ for correct plugin directory resolution (PRRT_kwDOO6Hdxs5uvKrg)
- Fix misleading trap registration comment in unraid-api-crawl.md and
add auth note to Apollo Studio URL (PRRT_kwDOO6Hdxs5uvO2t)
- Extract duplicated cleanup-with-error-handling in main.py into
_run_shutdown_cleanup() helper (PRRT_kwDOO6Hdxs5uvO3A)
- Add input validation to read-logs.sh to prevent GraphQL injection
via LOG_NAME and LINES parameters (PRRT_kwDOO6Hdxs5uvKrj)
- Move _ALLOWED_LOG_PREFIXES to module level (N806: constant naming)
- Use f-string conversion flag {e!s} instead of {str(e)} (RUF010)
- Fix import block sorting in both files (I001)
**Project Configuration:**
- Enhance pyproject.toml with comprehensive metadata, keywords, and classifiers
- Add LICENSE file (MIT) for proper open-source distribution
- Add PUBLISHING.md with comprehensive publishing guidelines
- Update .gitignore to exclude tool artifacts (.cache, .pytest_cache, .ruff_cache, .ty_cache)
- Ignore documentation working directories (.docs, .full-review, docs/plans, docs/sessions)
**Documentation:**
- Add extensive Unraid API research documentation
- API source code analysis and resolver mapping
- Competitive analysis and feature gap assessment
- Release notes analysis (7.0.0, 7.1.0, 7.2.0)
- Connect platform overview and remote access documentation
- Document known API patterns, limitations, and edge cases
**Testing & Code Quality:**
- Expand test coverage across all tool modules
- Add destructive action confirmation tests
- Improve test assertions and error case validation
- Refine type annotations for better static analysis
**Tool Improvements:**
- Enhance error handling consistency across all tools
- Improve type safety with explicit type annotations
- Refine GraphQL query construction patterns
- Better handling of optional parameters and edge cases
This commit prepares the project for v0.2.0 release with improved
metadata, comprehensive documentation, and enhanced code quality.
Co-authored-by: Claude <noreply@anthropic.com>
Refactor the entire tool layer to use the consolidated action pattern
(action: Literal[...] with QUERIES/MUTATIONS dicts). This reduces LLM
context from ~12k to ~5k tokens while adding ~60 new API capabilities.
New tools: unraid_info (19 actions), unraid_array (12), unraid_notifications (9),
unraid_users (8), unraid_keys (5). Rewritten: unraid_docker (15), unraid_vm (9),
unraid_storage (6), unraid_rclone (4), unraid_health (3).
Includes 129 tests across 10 test files, code review fixes for 16 issues
(severity ordering, PrefixedID regex, sensitive var redaction, etc.).
Removes tools/system.py (replaced by tools/info.py). Version bumped to 0.2.0.
- Remove array_status, system_info, notifications_overview, and parity_status resources
- Keep only logs_stream resource (unraid://logs/stream) which is working properly
- Update README.md with current resource documentation and modern docker compose syntax
- Fix import path issues that were causing subscription errors
- Update environment configuration examples
- Clean up subscription manager to only include working log streaming
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
