como um analista. Os únicos artefatos executáveis são scripts Python utilitários de pós-processamento (PDF). Para "rodar" o sistema, leia e siga `orchestrator.md`.
## Fluxo de execução
`SKILL.md` (entry point
verifiable citations.
**Default Format:** LaTeX with BibTeX citations unless otherwise requested.
**Quality Assurance:** Every PDF is automatically reviewed for formatting issues and iteratively improved until visually clean and professional.
**CRITICAL
fixes/tweaks, minor for new features/styles, major for breaking changes to the skill interface.
## PDF export
```bash
python3 scripts/export-pdf.py magazine.html
# Optional flags: --output filename.pdf --width 210 --height 297
# Requires: pip install
rules + bracket-placeholder scan on the ACTUAL outgoing .pdf/.docx
python scripts/lint-submission-file.py final/docx/ .docx final/pdf/ .pdf
```
`lint-submission-file.py` extracts text from the file that will actually be uploaded and applies the prose
candidate_profile/verified_facts.md` — real facts the user confirmed in
conversation that are not in the PDF yet. Equally source-of-truth. Update
it whenever the user confirms a new fact
facing docs (recap pages, deck narrative) go in your docs system. Binary files (xlsx, pdf, docx, pptx, images) go in your file storage. Emails in your email tool, calendar events
call remember(kind="document-note", title, key_points, summary, source_hint, category).
2. Only PDF and Word (.docx/.docm) accepted for file ingest. Images are described, not uploaded.
3. For image
step — the project is a pure-Python CLI. `pandoc` is an
optional system dep (PDF rendering only); `claude -p` is an optional binary
(enricher LLM path; falls back to rule
identity), plus a human-authored taste primer at `magic_trick.md`
and a `human/` directory for PDF materials the MCP intentionally ignores.
The goal: when an LLM helps build a website
source of truth. The pipeline produces **two artifacts**:
a print-ready 6×9" PDF (via Typst) and a reflowable EPUB 3 (via marked +
html-to-epub). A separate Typst document
/pdfs -o ./transcriptions --limit 5
# Test a single document before batching
make test-single PDF=./sample.pdf
# Theme consolidation (run after batch completes)
make consolidate DIR=./transcriptions # analyze only, no file
cmd/openrouter-test/main.go -test audiobuilder
go run cmd/openrouter-test/main.go -test audioformats
go run cmd/openrouter-test/main.go -test pdf
go run cmd/openrouter-test/main.go -test pdfengine
go run cmd/openrouter-test/main.go -test pdfannotations
go run cmd/openrouter-test/main.go -test base64pdf
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
Where does it go?
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
What should it contain?
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.md or AGENTS.md?
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.