report-workflowlisted

# Report Workflow Skill Operate the local deterministic pipeline: ```text prepare -> author -> validate -> render -> publish ``` The Python package does **not** call an LLM. It parses sources, builds an evidence ledger, checks agent-authored artifacts, renders DOCX, and packages outputs. You own judgment: claim selection, outlining, and drafting. Keep this file as the entry point and open a `reference/` file when a step needs detail. ## Reference Library Read the matching file only when the task reaches that step. Each file is one level deep and self-contained. | When you are… | Read | | --- | --- | | Checking deps, building `preflight_decisions` | [reference/setup-and-preflight.md](reference/setup-and-preflight.md) | | Choosing / understanding a profile | [reference/profiles.md](reference/profiles.md) | | Calling a tool and need its parameters | [reference/tools.md](reference/tools.md) | | Writing claims, outlines, drafts, citations | [reference/authoring.md](reference/authoring.md) | | Adding a figure, chart, schematic, or diagram | [reference/figures.md](reference/figures.md) | | Doing an `engineering_lab_report` | [reference/engineering-lab.md](reference/engineering-lab.md) | | Revising an existing DOCX | [reference/revision.md](reference/revision.md) | | Improving `report-workflow` itself | [reference/benchmarking.md](reference/benchmarking.md) | ## Core Model 1. **Prepare** — `start_report` reads sources, infers or accepts `report_profile`, and writes `report