Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Configure and optimize .geminiignore files for AI context window efficiency and token cost reduction (FinOps).
.claude/skills/geminiignore-finops/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | — | — |
| case-01 | ✗→✓ | ▲ Improved | — | — |
| case-07 | ✗→✓ | ▲ Improved | — | — |
| case-17 | ✓→✓ | = Same ✓ | — | — |
| case-18 | ✓→✓ | = Same ✓ | — | — |
A skill to construct, refine, and maintain high-performance .geminiignore files across diverse tech stacks. By filtering out machine-generated code, heavy logs, package locks, and binary assets, this skill optimizes the AI agent's context window, accelerates processing speed, and reduces token consumption costs (FinOps).
Detect the languages, frameworks, and dependency managers present in the project (e.g., Node.js, Python, PHP, Dart/Flutter, Rust).
.geminiignore FileCreate a .geminiignore file at the root of the active workspace. If one already exists, review it to add missing categories.
Add rules divided into the following categories to filter out unnecessary machine noise while keeping human-written code visible:
.DS_Store, Thumbs.db) and user-specific IDE caches (.idea/, .vscode/*, Xcode user data).node_modules/, vendor/) and giant machine-generated lock files (package-lock.json, yarn.lock, Cargo.lock, composer.lock).dist/, build/, .next/, .nuxt/)..tsbuildinfo, .vite/, .pytest_cache/, .eslintcache).*.png, *.pdf, *.mp4, *.woff2) to prevent triggering expensive vision/multimodal tokens.*.log) and SQL dumps or local SQLite DBs (*.sqlite, *.db).*.apk, *.ipa) and compiled binaries (*.class, *.pyc, *.dll).Verify that the AI can still see critical configuration blueprints (like .env.example, package.json, composer.json, pyproject.toml) but ignores the actual .env files and compilation artifacts.
.geminiignore TemplateHere is a recommended baseline configuration for a multi-language project:
ini# ============================================================================== # .geminiignore - BASELINE DE FINOPS E ARQUITETURA # ============================================================================== # 1. SISTEMA OPERACIONAL E IDEs .DS_Store Thumbs.db Desktop.ini $RECYCLE.BIN/ .vscode/* !.vscode/settings.json !.vscode/tasks.json !.vscode/launch.json .idea/ *.iml .gradle/ local.properties .history/ # 2. DEPENDÊNCIAS (ECONOMIA DE TOKENS EM LOCK FILES) node_modules/ package-lock.json yarn.lock pnpm-lock.yaml vendor/ composer.lock venv/ .venv/ env/ .env .env.* !.env.example poetry.lock Cargo.lock pubspec.lock # 3. BUILDS E EXPORTAÇÕES dist/ build/ out/ target/ .next/ .nuxt/ .output/ bin/ obj/ # 4. CACHES DE FRAMEWORKS .vite/ .parcel-cache/ .eslintcache .babel-cache/ .tsbuildinfo .turbo/ .pytest_cache/ .ruff_cache/ storage/framework/ storage/logs/ # 5. ASSETS BINÁRIOS E MULTIMÍDIA EXTREMOS *.png *.jpg *.jpeg *.gif *.webp *.svg *.ico *.psd *.fig *.pdf *.zip *.tar.gz *.woff *.woff2 *.ttf # 6. BANCOS DE DADOS E LOGS *.log *.db *.sqlite *.sqlite3 *.sql *.sql.gz # 7. ARQUIVOS COMPILADOS *.apk *.aab *.ipa *.jar *.class *.pyc __pycache__/ *.so *.dylib *.dll *.exe *.js.map *.css.map
package-lock.json, yarn.lock) contain thousands of lines of redundant package resolution trees. Ignoring them is the single largest FinOps win.package.json, composer.json, Cargo.toml, and pyproject.toml are NEVER ignored, as the AI needs them to understand dependencies.!.env.example alongside .env ignores so the AI understands configuration structure without exposing credentials.lib/ or app/ if they contain primary source code. Be specific (e.g., block vendor/bundle/ but not your actual code)..geminiignore file only affects AI tools parsing the workspace; it does not replace .gitignore for Git repository hosting.@context-optimization - Broad tactics for context window management.@clean-code - Architectural practices for clean, human-readable codebases.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
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 +14 percentage points is the difference between those two pass rates over the 21 comparable cases.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.