spinosa-writer

Generates numbered markdown research reports from evidence packets and goal artifacts.

7|Updated May 13, 2026
One-click install
npx skills add https://github.com/medialab/spinosa --skill spinosa-writer-medialab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: spinosa-writer
Source: https://github.com/medialab/spinosa/tree/main/workspace-template/.opencode/skills/spinosa-writer
Command: npx skills add https://github.com/medialab/spinosa --skill spinosa-writer-medialab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning raw search evidence into a coherent, citable research report is slow and error-prone. This Skill converts prior pipeline artifacts (evidence packets, analysis packets, goal artifacts) into structured, numbered markdown reports with source citations, without performing search or verification itself. ## Core Features & Use Cases - Structured Report Generation: Produces reports with Goal, TLDR, Report, Conclusions, Serendipity, and Reproducibility sections following a fixed template. - Sequential Numbering & Naming: Scans agent_reports/ for existing NN_*.md files and names new reports with an incremented number and topic slug. - Unicode Charts: Renders distribution bars, progress bars, status matrices, gauges, sparklines, and stacked bars for visual summaries. - Use Case: After a Searcher agent collects evidence from interview transcripts, invoke this Skill to write 03_coastal-erosion-normandy-interviews.md with inline source citations and a reproducibility table. ## Quick Start Ask the agent to write a numbered answer report from the current session's evidence packet and goal artifact into the agent_reports directory.

Frequently Asked Questions about spinosa-writer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a research report from collected evidence files?▼

Point the writer at the evidence packet and goal artifact paths from the session. It reads those files, structures findings into Goal, TLDR, Report, and Conclusions sections, and writes a numbered markdown report to the agent_reports directory.

How are report filenames and numbering determined?▼

The writer scans agent_reports for existing NN_*.md files, takes the highest number, and increments by one. The slug must state the research topic, such as 03_coastal-erosion-normandy-interviews.md, never generic names like NN_report.md.

Does the writer agent search or verify sources itself?▼

No. The writer only synthesizes artifacts produced by earlier pipeline steps. Searching is left to the Searcher agent and claim verification to the Verifier agent, keeping each step bounded and auditable.

What happens when the evidence set is very large?▼

When evidence exceeds roughly 300 lines or 50 sources, the main report includes only the top 10-20 sources and key patterns, while the full evidence set goes into an evidence appendix file referenced from the report.

What are the limitations of this report-writing approach?▼

The writer cannot invent evidence and depends entirely on the quality of upstream artifacts. If the evidence packet is incomplete or the goal artifact is missing, the resulting report will reflect those gaps rather than compensate for them.