How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
ur-grueCLAUDE.mdur-grue/autopunk-media-skills/CLAUDE.md
infographic-brief, │ │ │ data-table-formatter │ │ ├── investigation/ ← foi-data-request, scraper-brief, │ │ │ data-cleaning-brief, pdf-table-extractor-brief │ │ └── publishing/ ← methodology-explainer, data-corrections-writer, │ │ data-footnote-writer │ │ │ ├── research/ │ │ ├── people/ ← person
AlfonsSkillsCLAUDE.mdAlfonsSkills/SkillSync/docs/claude.md
Skills Anthropic provides pre-built Agent Skills for common document tasks (PowerPoint, Excel, Word, PDF), and you can create your own custom Skills. Both work the same way. Claude automatically
full-stack-bizCLAUDE.mdfull-stack-biz/claude-skills-toolkit/CLAUDE.md
docs → `references/api.md` ❌ **Vague or generic descriptions** - Don't write: "Process files" - DO write: "Process PDF files with OCR. Use when extracting text or analyzing documents. Supports encrypted PDFs." ❌ **Making design
geekjourneyxCLAUDE.mdgeekjourneyx/mind-skills/CLAUDE.md
Processes Excel files and generates reports. Use when working with spreadsheets." ✅ "Extracts text from PDF files. Use when user mentions PDFs or document extraction." ❌ "I can help you process Excel
zircoteCLAUDE.mdzircote/.claude/CLAUDE.md
repository structure, see `~/.claude/llms.txt`. ## Skills Quick Reference Skills are in `~/.claude/skills/`. Key categories: - **Documents**: `pdf`, `docx`, `xlsx`, `pptx` - **Media**: `ai-multimodal`, `media-processing`, `chrome-devtools` - **Development**: `frontend-development`, `backend-development
roomi-fieldsCLAUDE.mdroomi-fields/rtfm/.claude/CLAUDE.md
keys/arrays) │ ├── toml_parser.py # TOML (top-level tables; depends_on edges) │ ├── shell.py # Shell (function-aware) │ ├── pdf.py # PDF (pdftext/marker) │ ├── xml_legifrance.py # Legifrance XML │ ├── html_bofip.py # BOFiP HTML │ ├── sqlite_parser.py # SQLite DB (schema + ALL rows, streamed
vinnie357CLAUDE.mdvinnie357/claude-skills/CLAUDE.md
with formulas, charts, and formatting | | PowerPoint | `pptx` | Generate professional presentations with slides and layouts | | PDF | `pdf` | Produce formatted PDF documents | | Word | `docx` | Create rich-formatted Word documents | These skills demonstrate
AI-RiksarkivetCLAUDE.mdAI-Riksarkivet/ra-mcp/CLAUDE.md
handler scaffold `require_keyword` + `format_results`) that the 13 dataset libraries and the pdf-mcp guide search all reuse instead of each rolling their own search. Cross-cutting MCP helpers
zhurudongCLAUDE.mdzhurudong/andrej-karpathy-llm-wiki/templates/CLAUDE.md
架构 ``` . ├── raw/ # 不可变原始文章（LLM 只读，人类只在摄入时写入） │ ├── YYYY-MM-DD-标题.md │ └── assets/ # 原文附件（图片、PDF 等），按文章分子目录 │ └── / # 与 raw .md 同 stem（过长时可缩写） │ └── NN-描述.{png,jpg,pdf,...} ├── wiki
DataSift-Ty-PersonalCLAUDE.mdDataSift-Ty-Personal/SiftStack/CLAUDE.md
business lifecycle: 1. **Data Acquisition:** Web scraping tnpublicnotice.com (foreclosures, tax sales, probates), scanned PDF import, courthouse terminal photo import (probate, eviction, code violations, divorce), Dropbox auto-polling 2. **Enrichment Pipeline
ShieldNet-360CLAUDE.mdShieldNet-360/secure-vibe/dist/CLAUDE.md
redirect and scheme bypasses, parsers that fetch on their own (XXE, SVG, HTML-to-PDF), cloud metadata, and keeping the response from becoming an oracle. Use when fetching a client
beeblockCLAUDE.mdbeeblock/orkestrai/CLAUDE.md
policies, permissions, teams, middleware, rate limits, sessions, jobs, events, listeners, observers, cache, storage, search, PDF, and broadcasting through Svelar APIs instead of ad hoc implementations. ## Imports - Prefer app aliases such
BlockCLAUDE.mdblock/frfr/CLAUDE.md
verification. ## Stack - **Backend**: Go (REST API + SSE) - **Frontend**: React + TypeScript + Vite - **Desktop**: Electron (macOS) - **PDF extraction**: go-pdfium (WebAssembly, no Python) - **LLM**: Claude API ## Quick Start ```bash ./run.sh # Start backend
docxologyCLAUDE.mddocxology/template/CLAUDE.md
Overview This is a research project template with a test-driven development workflow, automated PDF generation, and multi-project support. It uses a two-layer architecture separating generic infrastructure (Layer
faugusteCLAUDE.mdfauguste/boondmanager-mcp-server/CLAUDE.md
apiDownload()` and, since issue #263, makes it **readable** by default (`mode: "text"`): a PDF goes through `unpdf` (a serverless pdf.js build, ~2 MB, no native binary), a DOCX through `src/services/document-text.ts
scvcoderCLAUDE.mdscvcoder/korean-law-alio-mcp/CLAUDE.md
기존 fallback 결과를 새 엔진으로 재생성 npm run alio:sync -- --docling-fallback # 스캔 이미지 PDF 를 docling OCR 로 복구 npm run alio:sync -- --concurrency 3 --keep-raw # 보고서형 공시
airano-irCLAUDE.mdairano-ir/skillhub/CLAUDE.md
Encoding ```typescript const encodedPath = id.split('/').map(encodeURIComponent).join('/'); ``` ### Meilisearch IDs Sanitized: `anthropics/skills/pdf` → `anthropics__skills__pdf` ## Environment Variables `DATABASE_URL`, `REDIS_URL`, `GITHUB_TOKEN`, `MEILI_URL` (optional), `MEILI_MASTER_KEY`, `GITHUB
anurmatovCLAUDE.mdanurmatov/phleet/CLAUDE.md
attachment: …]` only for `.jpg`/`.jpeg`/`.png` (what the provider paths consume); `[document attachment: …]` for `.pdf`; `[file attachment: …]` for everything else. - **Voice and video notes** are both transcribed through the whisper
applypackCLAUDE.mdapplypack/applypack/CLAUDE.md
through `app.request()`, the first run's POSTs, a cross-origin POST refused, one clean PDF render. A 500 anywhere fails the build. Run it locally on a throwaway database only
geobond13CLAUDE.mdgeobond13/fiction-forge/CLAUDE.md
02_Title.md`) - `reference/` — Story bible, characters, foreshadowing, continuity - `project.yaml` — All tool configuration - `output/` — Generated EPUB, PDF, HTML, scan reports ## Common Tasks ### Writing a new chapter ``` Write Chapter N: [Title
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
Build and test commands, the project's layout, conventions that aren't obvious from the code, and mistakes to avoid. Short files tend to work better than long ones.
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.
