Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use RepoPrompt CLI for token-efficient codebase exploration
.claude/skills/repoprompt/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 7% | 0% |
RepoPrompt is more token-efficient than raw file reads:
structure → signatures only (not full content)read --start-line --limit → slices instead of full filessearch --context-lines → relevant matches with contextbash# If installed to PATH (Settings → MCP Server → Install CLI to PATH) rp-cli -e 'command' # Or use the alias (configure in your shell) repoprompt_cli -e 'command'
bash# Full tree rp-cli -e 'tree' # Folders only rp-cli -e 'tree --mode folders' # Selected files only rp-cli -e 'tree --mode selected'
bash# Structure of specific paths rp-cli -e 'structure src/auth/' # Structure of selected files rp-cli -e 'structure --scope selected' # Limit results rp-cli -e 'structure src/ --max-results 10'
bash# Basic search rp-cli -e 'search "pattern"' # With context lines rp-cli -e 'search "error" --context-lines 3' # Filter by extension rp-cli -e 'search "TODO" --extensions .ts,.tsx' # Limit results rp-cli -e 'search "function" --max-results 20'
bash# Full file rp-cli -e 'read path/to/file.ts' # Line range (slice) rp-cli -e 'read path/to/file.ts --start-line 50 --limit 30' # Last N lines (tail) rp-cli -e 'read path/to/file.ts --start-line -20'
bash# Add files to selection rp-cli -e 'select add src/auth/' # Set selection (replace) rp-cli -e 'select set src/api/ src/types/' # Clear selection rp-cli -e 'select clear' # View current selection rp-cli -e 'select get'
bash# Get full context rp-cli -e 'context' # Specific includes rp-cli -e 'context --include prompt,selection,tree'
bash# Multiple operations rp-cli -e 'select set src/auth/ && structure --scope selected && context'
bash# List workspaces rp-cli -e 'workspace list' # List tabs rp-cli -e 'workspace tabs' # Switch workspace rp-cli -e 'workspace switch "ProjectName"'
bash# Send to chat rp-cli -e 'chat "How does the auth system work?"' # Plan mode rp-cli -e 'chat "Design a new feature" --mode plan'
bash# Auto-select relevant files for a task rp-cli -e 'builder "implement user authentication"'
bash# Quick operations without -e syntax rp-cli --workspace MyProject --select-set src/ --export-context ~/out.md rp-cli --chat "How does auth work?" rp-cli --builder "implement user authentication"
For repeatable workflows, save commands to a script:
bash# daily-export.rp workspace switch Frontend select set src/components/ context --all > ~/exports/frontend.md
Run with:
bashrp-cli --exec-file ~/scripts/daily-export.rp
| Flag | Purpose | |------|---------| | -e 'cmd' | Execute command(s) | | -w <id> | Target window ID | | -q | Quiet mode | | -d <cmd> | Detailed help for command | | --wait-for-server 5 | Wait for connection (scripts) |
For long-running operations like builder, use the async script:
bash# Start context builder async uv run python -m runtime.harness scripts/repoprompt_async.py \ --action start --task "understand the auth system" # With workspace switch uv run python -m runtime.harness scripts/repoprompt_async.py \ --action start --workspace "MyProject" --task "explore API patterns" # Check status uv run python -m runtime.harness scripts/repoprompt_async.py --action status # Get result when done uv run python -m runtime.harness scripts/repoprompt_async.py --action result # Kill if needed uv run python -m runtime.harness scripts/repoprompt_async.py --action kill
Requires RepoPrompt app running with MCP Server enabled.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | pass→pass | 5,844 | 2,581 | -56% | 1 | 1 | 0% | 1,017 | 1,693 | +66% | 0 | 0 | — |
case-01 | fail→pass | 5,139 | 2,159 | -58% | 1 | 1 | 0% | 939 | 1,629 | +73% | 0 | 0 | — |
case-02 | fail→pass | 6,872 | 2,259 | -67% | 1 | 1 | 0% | 1,316 | 1,716 | +30% | 0 | 0 | — |
case-03 | fail→pass | 6,345 | 2,841 | -55% | 1 | 1 | 0% | 1,181 | 1,794 | +52% | 0 | 0 | — |
case-04 | fail→fail | 6,259 | 2,160 | -65% | 1 | 1 | 0% | 1,277 | 1,675 | +31% | 0 | 0 | — |
case-05 | fail→pass | 9,769 | 1,837 | -81% | 1 | 1 | 0% | 1,660 | 1,591 | -4% | 0 | 0 | — |
case-06 | fail→pass | 8,125 | 1,574 | -81% | 1 | 1 | 0% | 1,418 | 1,512 | +7% | 0 | 0 | — |
case-07 | fail→pass | 5,460 | 1,969 | -64% | 1 | 1 | 0% | 905 | 1,569 | +73% | 0 | 0 | — |
case-08 | fail→pass | 6,924 | 1,640 | -76% | 1 | 1 | 0% | 1,281 | 1,502 | +17% | 0 | 0 | — |
case-14 | fail→pass | 5,470 | 1,676 | -69% | 1 | 1 | 0% | 980 | 1,547 | +58% | 0 | 0 | — |
case-09 | fail→pass | 23,497 | 1,539 | -93% | 1 | 1 | 0% | 704 | 1,460 | +107% | 0 | 0 | — |
case-10 | fail→pass | 7,297 | 1,204 | -84% | 1 | 1 | 0% | 1,319 | 1,418 | +8% | 0 | 0 | — |
case-11 | fail→pass | 6,873 | 2,373 | -65% | 1 | 1 | 0% | 1,412 | 1,623 | +15% | 0 | 0 | — |
case-12 | pass→pass | 4,778 | 1,572 | -67% | 1 | 1 | 0% | 815 | 1,504 | +85% | 0 | 0 | — |
case-15 | pass→pass | 20,445 | 2,320 | -89% | 1 | 1 | 0% | 1,427 | 1,712 | +20% | 0 | 0 | — |
case-16 | fail→pass | 6,934 | 2,406 | -65% | 1 | 1 | 0% | 1,330 | 1,645 | +24% | 0 | 0 | — |
case-17 | pass→pass | 3,884 | 14,253 | +267% | 1 | 1 | 0% | 703 | 1,540 | +119% | 0 | 0 | — |
case-18 | fail→pass | 8,862 | 2,012 | -77% | 1 | 1 | 0% | 1,519 | 1,615 | +6% | 0 | 0 | — |
case-19 | fail→pass | 10,185 | 2,427 | -76% | 1 | 1 | 0% | 1,820 | 1,758 | -3% | 0 | 0 | — |
case-20 | fail→pass | 7,807 | 1,576 | -80% | 1 | 1 | 0% | 1,511 | 1,543 | +2% | 0 | 0 | — |
case-21 | fail→pass | 2,436 | 1,593 | -35% | 1 | 1 | 0% | 413 | 1,504 | +264% | 0 | 0 | — |
case-22 | pass→pass | 2,401 | 2,444 | +2% | 1 | 1 | 0% | 419 | 1,643 | +292% | 0 | 0 | — |
case-23 | pass→pass | 6,802 | 3,214 | -53% | 1 | 1 | 0% | 1,279 | 1,819 | +42% | 0 | 0 | — |
case-24 | pass→pass | 4,855 | 4,427 | -9% | 1 | 1 | 0% | 796 | 1,904 | +139% | 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. 24 cases were attempted. The headline lift of +67 percentage points is the difference between those two pass rates over the 24 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/31/2026 | +88% |
Other measured skills in the registry, with their headline benchmark lift.