Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Core Smart Ralph skill defining common arguments, execution modes, and shared behaviors across all Ralph plugins.
.claude/skills/tzachbon-smart-ralph/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -39% | 0% |
Core skill for all Ralph plugins. Defines common arguments, execution modes, shared behaviors, and coordinator delegation rules.
All Ralph commands support these standard arguments:
| Argument | Short | Description | Default | |----------|-------|-------------|---------| | --quick | -q | Skip interactive phases, auto-generate artifacts, start execution immediately | false | | --commit | -c | Commit and push spec/feature files after generation | true (normal), false (quick) | | --no-commit | | Explicitly disable committing files | - | | --max-task-iterations | -m | Max retries per failed task before stopping | 5 | | --fresh | -f | Force new spec/feature, overwrite if exists | false |
Argument precedence: --no-commit > --commit > mode default.
awaitingApproval: true--quick)--commit to override)All Ralph plugins use .ralph-state.json for execution state. See references/state-file-schema.md for full schema.
Key fields: phase, taskIndex, totalTasks, taskIteration, maxTaskIterations, awaitingApproval.
When commitSpec is true:
chore(<plugin>): commit spec files before implementationWhen commitSpec is false:
Ralph Specum v3.0.0+ has a self-contained execution loop via the stop-hook. No external dependencies required.
Key signals:
TASK_COMPLETE - executor finished taskALL_TASKS_COMPLETE - coordinator ends loopWhen taskIteration > maxTaskIterations: block task, suggest manual intervention.
If state file missing/invalid: output error, suggest re-running implement command.
All Ralph plugins follow consistent branch strategy:
The main agent is a coordinator, not an implementer. Delegate all work to subagents.
| Work Type | Delegate To | |-----------|-------------| | Research | Research Team (parallel teammates) | | Requirements | product-manager subagent | | Design | architect-reviewer subagent | | Task planning | task-planner subagent | | Task execution | spec-executor subagent |
Quick mode still requires delegation.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,654 | 6,547 | -55% | 1 | 1 | 0% | 2,300 | 1,928 | -16% | 0 | 0 | — |
case-02 | fail→pass | 21,418 | 5,831 | -73% | 1 | 1 | 0% | 3,072 | 1,722 | -44% | 0 | 0 | — |
case-03 | pass→pass | 16,739 | 10,395 | -38% | 1 | 1 | 0% | 2,694 | 2,585 | -4% | 0 | 0 | — |
case-04 | fail→pass | 13,056 | 2,640 | -80% | 1 | 1 | 0% | 1,777 | 1,164 | -34% | 0 | 0 | — |
case-05 | pass→pass | 8,142 | 1,996 | -75% | 1 | 1 | 0% | 1,050 | 1,084 | +3% | 0 | 0 | — |
case-06 | fail→pass | 12,949 | 2,224 | -83% | 1 | 1 | 0% | 1,762 | 1,166 | -34% | 0 | 0 | — |
case-07 | pass→pass | 12,692 | 4,051 | -68% | 1 | 1 | 0% | 1,769 | 1,431 | -19% | 0 | 0 | — |
case-08 | pass→pass | 12,163 | 1,862 | -85% | 1 | 1 | 0% | 1,827 | 1,091 | -40% | 0 | 0 | — |
case-09 | pass→pass | 11,112 | 1,652 | -85% | 1 | 1 | 0% | 1,460 | 1,039 | -29% | 0 | 0 | — |
case-14 | fail→pass | 14,537 | 2,640 | -82% | 1 | 1 | 0% | 2,022 | 1,232 | -39% | 0 | 0 | — |
case-10 | fail→pass | 16,056 | 1,622 | -90% | 1 | 1 | 0% | 2,408 | 1,038 | -57% | 0 | 0 | — |
case-11 | fail→pass | 9,857 | 1,980 | -80% | 1 | 1 | 0% | 1,471 | 1,121 | -24% | 0 | 0 | — |
case-12 | pass→pass | 15,880 | 5,447 | -66% | 1 | 1 | 0% | 2,177 | 1,674 | -23% | 0 | 0 | — |
case-13 | fail→pass | 12,547 | 4,386 | -65% | 1 | 1 | 0% | 1,799 | 1,514 | -16% | 0 | 0 | — |
case-15 | fail→pass | 16,290 | 2,202 | -86% | 1 | 1 | 0% | 1,509 | 1,176 | -22% | 0 | 0 | — |
case-16 | pass→pass | 9,974 | 3,068 | -69% | 1 | 1 | 0% | 1,464 | 1,310 | -11% | 0 | 0 | — |
case-17 | fail→pass | 7,807 | 1,534 | -80% | 1 | 1 | 0% | 1,085 | 971 | -11% | 0 | 0 | — |
case-18 | fail→pass | 9,938 | 2,252 | -77% | 1 | 1 | 0% | 1,569 | 1,147 | -27% | 0 | 0 | — |
case-19 | fail→pass | 14,865 | 2,908 | -80% | 1 | 1 | 0% | 1,985 | 1,263 | -36% | 0 | 0 | — |
case-20 | pass→pass | 14,297 | 13,004 | -9% | 1 | 1 | 0% | 2,235 | 2,914 | +30% | 0 | 0 | — |
case-21 | pass→pass | 10,369 | 5,722 | -45% | 1 | 1 | 0% | 1,675 | 1,744 | +4% | 0 | 0 | — |
case-22 | pass→pass | 13,763 | 12,695 | -8% | 1 | 1 | 0% | 2,542 | 3,224 | +27% | 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. The headline lift of +55 percentage points is the difference between those two pass rates over the 22 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.