Price terms as published
Each line is the vendor’s own wording, split from the price capture checked on 2026-09-29. Nothing here is rounded, converted or re-worded — the full sentences stay in the review below. Lines that carry only a price continue the plan named on the line above, in the vendor’s own order.
| # | Plan / term | Checked |
|---|---|---|
| 1 | Free (open source, MIT) | 2026-09-29 |
One-liner: Huashu-Report is an agent skill that makes an AI work through four roles — researcher, editor, information designer, and data-visualization specialist — when producing industry reports, white papers, annual surveys, or arXiv-style papers.
Verified facts (checked 2026-09-29)
| Attribute | Value | Source |
|---|---|---|
| GitHub stars | 428 | alchaincyf/huashu-report via GitHub API |
| License | MIT | GitHub API |
| Created | 2026-08-31 | GitHub API |
| Last push | 2026-09-14 | GitHub API |
| Format | Plain-text skill plus three Python files | README |
| Evidence corpus | 43 sample entries; quantified anatomy of 41 reports | README / corpus/ |
How the skill was built
The author’s stated method: download top-tier institutional AI reports — Stanford HAI, McKinsey, BCG, OpenAI, Anthropic, PwC, Deloitte, World Bank, Reuters Institute, and others — and reverse-engineer what their structure, citation labeling, chart titles, and color use have in common.
The corpus is documented in the repository (corpus/samples.md with institution, date, page count, official link, and license status; corpus/anatomy.json with the quantified measurements). The PDF files themselves are deliberately not redistributed.
Three quantified findings cited in the README
- Near-black body text dominates. In 40 of 41 reports (98%), body text is near-black; brand color covers a median of only about 2.8% of body characters — brand color is reserved for charts, section markers, and emphasized numbers.
- Chart titles state conclusions, not topics. “The employment gap for young workers keeps widening” rather than “Employment changes by age group” — with research-report-style titles as the documented exception.
- Top reports pre-refute themselves. The example given is Stanford’s “Canaries” report downgrading its findings to descriptive rather than causal indicators — a section that becomes the anchor others cite.
These are presented as measured patterns from the corpus, which is a stronger basis than generic “write like a consultant” advice.
What is inside
The repository includes a SKILL.md with three required kickoff questions and a selection table for six report prototypes; reference files on structure, prose, research depth, visual systems, chart patterns, and a production pipeline; and assets for inline SVG charts and rendering. One stated convenience: the two Python files in assets/ are meant to be copied and configured rather than rebuilt.
A single editing test is given in the README: “If I delete this sentence, will the reader’s judgment about the research conclusion change? If not, it is a work log — delete it.”
Caveats
- Small, individual project. 428 stars and a single maintainer. Treat it as a focused craft asset, not a supported platform.
- A style skill cannot manufacture research quality. It organizes and formats work; garbage or unverified inputs still produce a polished-but-wrong report. Source verification remains the operator’s job.
- The corpus has a domain tilt. The sample set skews toward AI and institutional research; its quantified conventions may transfer imperfectly to, for example, consumer marketing or technical API documentation.
- “Institutional-grade” describes the structure, not automatic credibility with readers — a report earns trust from its evidence.
Who should use it
Analysts, founders, and researchers who use an agent to write substantial reports and want an enforced structure with explicit evidence and visual conventions instead of one long undifferentiated draft. Writers who already have strong report discipline may find it most useful as a checklist and a source of quantified baselines.
Repository facts and corpus claims verified against GitHub and the project README on 2026-09-29. This is a draft pending human editorial review.