| name | write-professional-readme |
| description | Use when creating, rewriting, auditing, or localizing a professional README.md or project landing document for a repository, library, API, CLI, web app, data project, or generated problem corpus. Trigger for requests about README generation, README polish, installation/usage docs, quick starts, badges, feature summaries, project structure, contribution sections, or making documentation clearer and more credible. |
Write Professional README
Create README files that are accurate to the codebase, useful to first-time users, and polished
enough for public release. Prefer evidence from repository files over generic filler.
Workflow
- Inspect the repository before drafting: read existing
README.md, package/build files, entry
points, tests, examples, docs, and public API surfaces.
- Identify the audience and purpose: user-facing product, developer library, CLI tool, API service,
research/demo project, internal tool, or problem corpus.
- Draft around real capabilities only. If a feature is inferred, mark it as inferred in your
reasoning and verify through code or tests before presenting it as fact.
- Prefer a concise, scannable README structure. Put the most useful setup and usage path near the
top.
- Preserve repository-specific names, commands, endpoint spellings, environment variables, and file
layout. Do not invent branding, badges, screenshots, licenses, or deployment targets.
- Validate commands when feasible with local, non-destructive checks. If commands require secrets,
network, live services, or destructive state changes, document them without running them.
- When editing an existing README, keep valuable existing content and remove stale, duplicate, or
marketing-heavy text.
- Read
references/readme-quality-checklist.md before finalizing a new README or a substantial
rewrite.
Recommended Structure
Choose only the sections that fit the project:
- Project title and one-sentence value proposition
- What it does, with concrete capabilities
- Requirements or prerequisites
- Installation or setup
- Quick start
- Configuration, including
.env or config file notes without exposing secrets
- Common commands
- API, CLI, or UI usage examples
- Project structure
- Testing and validation
- Generated output shape or important artifacts
- Development notes, contribution workflow, and troubleshooting
- License, only when a license file or explicit project metadata exists
Writing Standards
- Lead with utility, not hype.
- Use exact commands in fenced code blocks.
- Keep headings predictable and skimmable.
- Explain configuration boundaries clearly, especially secret-bearing files.
- Include examples that match the repository's actual interface.
- State limitations honestly when setup requires external services, credentials, or unfinished work.
- Match the user's requested language. If no language is specified, follow the repository's existing
README language; otherwise use clear English for international developer projects.
- Avoid boilerplate sections that cannot be supported by repository evidence.
Repository-Specific Guardrails
For this ACMFlow repository:
- Treat
config.ini as secret-bearing. Mention how to configure it, but do not print credentials.
- Preserve the
/promblemserve endpoint spelling unless the user explicitly asks to rename it.
- Describe generated problem bundles using the real output shape under
problem/normal/<title>/.
- Prefer local validation commands such as
python tests/parser_smoke.py,
python tests/test_pairwise.py, and the repository-local skill scripts when relevant.
- Keep root-level entrypoint guidance centered on
python main.py.
Output Modes
When the user asks to create or update a README, edit the file directly unless they only requested a
draft. When the user asks for a review, provide findings first with file/line references and suggest
specific README changes.
References
- Read
references/readme-quality-checklist.md before publishing a new README or major rewrite.