Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →To connect Rosetta with Grid Dynamics SpecFlow MCP; only when SpecFlow is mentioned and the MCP is installed.
.claude/skills/griddynamics-specflow-use/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -20% | 0% |
This skill helps you drive Grid Dynamics SpecFlow from a local workspace via SpecFlow MCP. Rosetta stays local; SpecFlow runs everything remotely. The skill loads project context from gain.json and guides the user through SpecFlow MCP tool usage.
SpecFlow MCP is the only user interface to SpecFlow: you order work and download outputs. You do not connect to or operate on the remote workspaces directly.
When the user mentions SpecFlow or works in a SpecFlow-enabled project:
gain.json at the workspace root (and up to 2 parents). If missing, SpecFlow MCP will create it on the first tool call (together with specflow_session.json).references/specflow-mcp-tools.md.Check for gain.json at the current workspace root. If not found, check parent directories up to 2 levels. If still missing, tell the user this workspace is not yet initialized for SpecFlow — the first SpecFlow MCP tool call will create both gain.json and specflow_session.json.
When gain.json is found, present a compact summary:
## SpecFlow Project Context
**Description**: [gain.json.description]
**Services**: specflow — [gain.json.servicesDescription.specflow]
rosetta — [gain.json.servicesDescription.rosetta]
**Supported Coding Agents**: [gain.json.codingAgents]
**Versions**:
- rosetta: [gain.json.versions.rosetta]
- specflow: [gain.json.versions.specflow]
**Vocabulary overrides** (if present in gain.json.vocabulary):
- (See references/specflow-vocabulary.md for defaults.)Compare gain.json.versions.rosetta with the current Rosetta version:
APPLY SKILL FILE references/specflow-mcp-tools.md to route the user to the correct phase (pregeneration, generation, or post-run).
When SpecFlow returns a review report and the user wants to fix their specs before another iteration, delegate to Rosetta core:
skills/requirements-authoringprompts/requirements-authoring-flow.prompt.mdThat skill handles iterative requirement updates with explicit user approval — matching the "consolidate the review report into updated specs" loop.
Read as needed:
references/specflow-mcp-tools.md — Phase routing and post-run skill bootstrap.references/specflow-vocabulary.md — Status values and checkpoint gates users encounter in responses.references/specflow-schema.md — Session files, version channels, and status enum quick lookup.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,619 | 4,005 | -62% | 1 | 1 | 0% | 1,694 | 1,309 | -23% | 0 | 0 | — |
case-11 | fail→pass | 10,214 | 3,106 | -70% | 1 | 1 | 0% | 1,471 | 1,289 | -12% | 0 | 0 | — |
case-02 | pass→pass | 5,435 | 5,038 | -7% | 1 | 1 | 0% | 1,069 | 1,541 | +44% | 0 | 0 | — |
case-03 | fail→pass | 15,826 | 13,143 | -17% | 1 | 1 | 0% | 2,803 | 2,978 | +6% | 0 | 0 | — |
case-04 | fail→fail | 6,019 | 1,857 | -69% | 1 | 1 | 0% | 259 | 999 | +286% | 0 | 0 | — |
case-05 | fail→fail | 6,238 | 3,339 | -46% | 1 | 1 | 0% | 439 | 1,117 | +154% | 0 | 0 | — |
case-06 | fail→fail | 8,945 | 3,484 | -61% | 1 | 1 | 0% | 1,381 | 1,172 | -15% | 0 | 0 | — |
case-07 | fail→pass | 9,369 | 5,895 | -37% | 1 | 1 | 0% | 1,532 | 1,722 | +12% | 0 | 0 | — |
case-08 | fail→pass | 10,215 | 3,716 | -64% | 1 | 1 | 0% | 1,645 | 1,319 | -20% | 0 | 0 | — |
case-09 | pass→pass | 9,269 | 3,315 | -64% | 1 | 1 | 0% | 1,520 | 1,213 | -20% | 0 | 0 | — |
case-10 | pass→pass | 7,514 | 2,998 | -60% | 1 | 1 | 0% | 1,216 | 1,213 | -0% | 0 | 0 | — |
case-12 | fail→fail | 8,923 | 3,087 | -65% | 1 | 1 | 0% | 1,310 | 1,201 | -8% | 0 | 0 | — |
case-13 | fail→pass | 18,701 | 3,430 | -82% | 1 | 1 | 0% | 880 | 1,317 | +50% | 0 | 0 | — |
case-14 | fail→pass | 11,240 | 2,956 | -74% | 1 | 1 | 0% | 1,796 | 1,202 | -33% | 0 | 0 | — |
case-15 | fail→pass | 13,200 | 3,058 | -77% | 1 | 1 | 0% | 1,995 | 1,197 | -40% | 0 | 0 | — |
case-16 | fail→pass | 15,116 | 2,032 | -87% | 1 | 1 | 0% | 2,661 | 1,037 | -61% | 0 | 0 | — |
case-17 | fail→pass | 9,739 | 1,702 | -83% | 1 | 1 | 0% | 1,425 | 970 | -32% | 0 | 0 | — |
case-18 | fail→pass | 12,783 | 3,124 | -76% | 1 | 1 | 0% | 2,071 | 1,225 | -41% | 0 | 0 | — |
case-19 | fail→pass | 10,221 | 3,343 | -67% | 1 | 1 | 0% | 1,474 | 1,275 | -14% | 0 | 0 | — |
case-20 | fail→pass | 9,459 | 2,491 | -74% | 1 | 1 | 0% | 1,409 | 1,168 | -17% | 0 | 0 | — |
case-21 | fail→fail | 9,351 | 4,881 | -48% | 1 | 1 | 0% | 1,350 | 1,484 | +10% | 0 | 0 | — |
case-22 | fail→pass | 15,432 | 4,831 | -69% | 1 | 1 | 0% | 2,268 | 1,505 | -34% | 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 21 counted toward the lift figure. The other 1 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 +64 percentage points is the difference between those two pass rates over the 21 comparable cases.
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.