Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Map code questions to the optimal tldr command by detecting intent and routing to the right analysis layer.
.claude/skills/tldr-router/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -42% | 0% |
Maps questions to the optimal tldr command. Use this to pick the right layer.
bashtldr tree . --ext .py # File overview tldr structure src/ --lang python # Function/class overview
Use: Starting exploration, orientation
bashtldr context <function> --project . --depth 2 tldr calls src/
Use: Understanding architecture, finding entry points
bashtldr cfg <file> <function>
Use: Identifying refactoring candidates, understanding difficulty
bashtldr dfg <file> <function>
Use: Debugging, understanding data flow
bashtldr slice <file> <function> <line>
Use: Impact analysis, safe refactoring
bashtldr search "pattern" src/
Use: Finding code, structural search
START
│
├─► "What exists?" ──► tree / structure
│
├─► "How does X connect?" ──► context / calls
│
├─► "Why is X complex?" ──► cfg
│
├─► "Where does Y flow?" ──► dfg
│
├─► "What depends on Z?" ──► slice
│
└─► "Find something" ──► search| Intent | Keywords | Layer | |--------|----------|-------| | Navigation | "what", "where", "find", "exists" | tree, structure, search | | Architecture | "calls", "uses", "connects", "depends" | context, calls | | Complexity | "complex", "refactor", "branches", "paths" | cfg | | Data Flow | "variable", "value", "assigned", "comes from" | dfg | | Impact | "affects", "changes", "slice", "dependencies" | slice/pdg | | Debug | "bug", "error", "investigate", "broken" | cfg + dfg + context |
The tldr-read-enforcer and tldr-context-inject hooks automatically:
You don't need to manually run these commands - the hooks do it for you.
If you need a specific layer the hooks didn't provide:
bash# Force specific analysis tldr cfg path/to/file.py function_name tldr dfg path/to/file.py function_name tldr slice path/to/file.py function_name 42
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,811 | 2,342 | -85% | 1 | 1 | 0% | 2,703 | 1,073 | -60% | 0 | 0 | — |
case-02 | fail→pass | 8,755 | 2,135 | -76% | 1 | 1 | 0% | 1,738 | 1,034 | -41% | 0 | 0 | — |
case-03 | fail→pass | 17,091 | 3,500 | -80% | 1 | 1 | 0% | 3,120 | 1,045 | -67% | 0 | 0 | — |
case-04 | pass→pass | 5,038 | 5,106 | +1% | 1 | 1 | 0% | 898 | 1,620 | +80% | 0 | 0 | — |
case-05 | pass→pass | 4,192 | 3,300 | -21% | 1 | 1 | 0% | 762 | 1,195 | +57% | 0 | 0 | — |
case-06 | pass→pass | 3,954 | 4,141 | +5% | 1 | 1 | 0% | 690 | 1,291 | +87% | 0 | 0 | — |
case-07 | fail→pass | 10,113 | 2,153 | -79% | 1 | 1 | 0% | 1,945 | 1,069 | -45% | 0 | 0 | — |
case-08 | fail→pass | 10,871 | 2,919 | -73% | 1 | 1 | 0% | 2,009 | 1,173 | -42% | 0 | 0 | — |
case-09 | fail→pass | 12,206 | 2,266 | -81% | 1 | 1 | 0% | 1,979 | 999 | -50% | 0 | 0 | — |
case-10 | fail→pass | 10,927 | 2,643 | -76% | 1 | 1 | 0% | 2,153 | 1,110 | -48% | 0 | 0 | — |
case-11 | fail→pass | 6,168 | 1,657 | -73% | 1 | 1 | 0% | 1,127 | 967 | -14% | 0 | 0 | — |
case-12 | fail→pass | 11,402 | 2,207 | -81% | 1 | 1 | 0% | 1,905 | 1,043 | -45% | 0 | 0 | — |
case-13 | fail→pass | 19,424 | 3,137 | -84% | 1 | 1 | 0% | 2,980 | 975 | -67% | 0 | 0 | — |
case-14 | fail→pass | 25,181 | 4,948 | -80% | 1 | 1 | 0% | 2,050 | 991 | -52% | 0 | 0 | — |
case-15 | fail→pass | 11,301 | 3,826 | -66% | 1 | 1 | 0% | 1,825 | 1,002 | -45% | 0 | 0 | — |
case-16 | pass→pass | 7,430 | 3,248 | -56% | 1 | 1 | 0% | 1,172 | 1,314 | +12% | 0 | 0 | — |
case-17 | fail→pass | 6,856 | 2,113 | -69% | 1 | 1 | 0% | 1,113 | 1,007 | -10% | 0 | 0 | — |
case-18 | fail→pass | 6,823 | 2,339 | -66% | 1 | 1 | 0% | 1,069 | 1,060 | -1% | 0 | 0 | — |
case-19 | pass→pass | 9,241 | 3,697 | -60% | 1 | 1 | 0% | 1,566 | 1,174 | -25% | 0 | 0 | — |
case-20 | fail→pass | 6,679 | 2,121 | -68% | 1 | 1 | 0% | 1,109 | 1,043 | -6% | 0 | 0 | — |
case-21 | fail→pass | 10,404 | 2,572 | -75% | 1 | 1 | 0% | 1,732 | 1,107 | -36% | 0 | 0 | — |
case-22 | fail→pass | 7,901 | 1,557 | -80% | 1 | 1 | 0% | 1,312 | 950 | -28% | 0 | 0 | — |
case-23 | fail→pass | 10,726 | 2,797 | -74% | 1 | 1 | 0% | 1,887 | 1,007 | -47% | 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. 23 cases were attempted, and 22 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 +78 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/29/2026 | +80% |
Other measured skills in the registry, with their headline benchmark lift.