Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use the OpenProject Codex plugin to inspect and manage work in Codex through the OpenProject API.
.claude/skills/hashgraph-online-openproject-codex/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 92% | 12 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 41% | 0% |
Use this plugin when the user wants to work with OpenProject from Codex instead of using the OpenProject UI.
The plugin is intended to cover common project-management actions directly and then fall back to the generic authenticated API tool for any OpenProject module that is not yet represented by a dedicated MCP tool.
OPENPROJECT_DEFAULT_PROJECT when present, unless the user explicitly asks for another project.openproject_connection_status first if credentials or connectivity are uncertain.openproject_setup_connection to capture the base URL and API token directly from chat before attempting other actions.openproject_test_connection or openproject_whoami after setup when you need a quick verification step.OPENPROJECT_UI_USERNAME and OPENPROJECT_UI_PASSWORD are configured.openproject_list_types, openproject_list_statuses, openproject_list_priorities, and openproject_list_project_assignees before write operations when a value needs to be discovered.assignee_name, responsible_name, status_name, priority_name, and customer on openproject_create_work_package and openproject_update_work_package when the user provides human-readable values.openproject_list_custom_field_options or openproject_resolve_custom_field_option for Customer or other list-style custom fields instead of using the browser UI.openproject_my_work for “my items” or “items assigned to me” requests.openproject_call_api against the relevant /api/v3/... endpoint instead of sending the user back to OpenProject UI./mcp endpoint is read-only, so write operations should continue to use this plugin's API tools.OPENPROJECT_API_TOKEN is the preferred credential.OPENPROJECT_API_TOKEN_FILE is supported for local secret-file storage and is the recommended Codex desktop setup.OPENPROJECT_BASIC_API_TOKEN is supported as a fallback for legacy Basic auth with the apikey user.OPENPROJECT_BASIC_API_TOKEN_FILE is supported for the same legacy Basic-auth fallback.OPENPROJECT_BASE_URL is required.OPENPROJECT_DEFAULT_PROJECT is optional.OPENPROJECT_UI_USERNAME and OPENPROJECT_UI_PASSWORD are required for UI-session API fallback and the boards/wiki/meetings UI-backed tools.The plugin can now persist this configuration locally through openproject_setup_connection, so the user does not have to edit .mcp.json manually on first use.
openproject_call_api for unsupported modules or advanced API workflows.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,228 | 15,619 | +69% | 1 | 1 | 0% | 571 | 1,347 | +136% | 0 | 0 | — |
case-02 | fail→fail | 18,326 | 7,132 | -61% | 1 | 1 | 0% | 560 | 1,428 | +155% | 0 | 0 | — |
case-03 | fail→fail | 14,418 | 17,831 | +24% | 1 | 1 | 0% | 1,572 | 1,445 | -8% | 0 | 0 | — |
case-04 | fail→fail | 14,357 | 10,592 | -26% | 1 | 1 | 0% | 1,455 | 1,433 | -2% | 0 | 0 | — |
case-05 | fail→fail | 11,308 | 11,394 | +1% | 1 | 1 | 0% | 718 | 1,697 | +136% | 0 | 0 | — |
case-06 | pass→pass | 11,308 | 10,021 | -11% | 1 | 1 | 0% | 1,903 | 1,766 | -7% | 0 | 0 | — |
case-07 | fail→pass | 5,928 | 9,462 | +60% | 1 | 1 | 0% | 941 | 1,646 | +75% | 0 | 0 | — |
case-08 | fail→pass | 13,168 | 10,595 | -20% | 1 | 1 | 0% | 2,061 | 1,928 | -6% | 0 | 0 | — |
case-09 | fail→pass | 32,131 | 10,511 | -67% | 1 | 1 | 0% | 2,270 | 1,846 | -19% | 0 | 0 | — |
case-10 | fail→fail | 23,972 | 12,035 | -50% | 1 | 1 | 0% | 1,607 | 1,314 | -18% | 0 | 0 | — |
case-11 | fail→fail | 10,843 | 9,505 | -12% | 1 | 1 | 0% | 1,081 | 1,675 | +55% | 0 | 0 | — |
case-12 | pass→fail | 11,180 | 9,335 | -17% | 1 | 1 | 0% | 1,722 | 1,278 | -26% | 0 | 0 | — |
case-13 | fail→pass | 34,737 | 11,487 | -67% | 1 | 1 | 0% | 2,021 | 1,556 | -23% | 0 | 0 | — |
case-14 | fail→fail | 11,283 | 7,974 | -29% | 1 | 1 | 0% | 2,051 | 1,285 | -37% | 0 | 0 | — |
case-15 | fail→fail | 15,079 | 7,439 | -51% | 1 | 1 | 0% | 1,654 | 1,313 | -21% | 0 | 0 | — |
case-16 | fail→fail | 4,324 | 15,835 | +266% | 1 | 1 | 0% | 594 | 1,277 | +115% | 0 | 0 | — |
case-17 | fail→pass | 10,208 | 12,830 | +26% | 1 | 1 | 0% | 1,668 | 2,346 | +41% | 0 | 0 | — |
case-18 | fail→pass | 9,862 | 8,947 | -9% | 1 | 1 | 0% | 1,461 | 1,481 | +1% | 0 | 0 | — |
case-19 | fail→fail | 13,229 | 10,983 | -17% | 1 | 1 | 0% | 1,520 | 1,137 | -25% | 0 | 0 | — |
case-20 | pass→pass | 15,468 | 13,862 | -10% | 1 | 1 | 0% | 1,823 | 2,680 | +47% | 0 | 0 | — |
case-21 | pass→pass | 14,002 | 10,004 | -29% | 1 | 1 | 0% | 2,419 | 3,078 | +27% | 0 | 0 | — |
case-22 | pass→pass | 9,113 | 5,839 | -36% | 1 | 1 | 0% | 1,248 | 1,985 | +59% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 14 counted toward the lift figure. The other 8 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +23 percentage points is the difference between those two pass rates over the 14 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.