Reader's Seat

Turn scattered source material into complete documents readers can understand, verify, and act on.

中文说明 · Prompt examples · Install

Information-rich sources can still leave the reader to reconstruct the point: which result matters, what supports it, and what decision follows. Reader's Seat turns that material into a complete document with the conclusion, evidence, limits, and next step visible, without inventing missing facts.

① Original webpage on the left → ② Reader's Seat output on the right

_{The left panel uses a real continuous screenshot excerpt from the live Artificial Analysis model comparison; the right panel synthesizes all 11 comparison sections. The preview is a dated snapshot and rankings will change.}

Install

Ask Your Agent

Send this instruction to an agent that can install skills:

Install Reader's Seat from https://github.com/BobbyYue/reader-seat.
Use the skill installer supported by this agent, keep the folder name reader-seat,
then confirm that SKILL.md can be loaded. Do not overwrite an existing installation without asking.

Clone Manually

Clone the repository into the skill directory used by your agent:

git clone https://github.com/BobbyYue/reader-seat.git /path/to/your-agent/skills/reader-seat

Then start a new agent session or ask the agent to load reader-seat/SKILL.md explicitly. Skill discovery differs between agents.

What You Can Make

Start with	Get a complete document
Weekly reports, milestones, meeting notes, issue lists	Project status brief with the current conclusion, progress, blockers, decisions, and next steps
Metric definitions, query results, charts, methodology notes	Analysis report with the finding and magnitude, supporting evidence, data limits, alternative explanations, and next validation
Requirements, constraints, candidate architectures, benchmarks	Technical decision document with the decision, option comparison, trade-offs, implementation impact, rollback, and unresolved risks

It also supports news and industry briefs, product documents, retrospectives, performance summaries, SOPs, runbooks, and help articles.

Use It

Name the material, reader, and job the document must do. For example:

Use Reader's Seat to turn these weekly reports, milestone plan, meeting notes,
and issue list into a complete project status report for the leadership team.
Make the current conclusion, schedule risk, supporting evidence, decisions needed,
and next steps easy to find. Preserve unknowns instead of filling them in.

See ready-to-use prompts for project progress, data analysis, and technical route documents.

What It Protects

• Meaning: facts, quantities, definitions, ownership, commitments, and uncertainty remain intact.
• Judgment: conclusions stay connected to evidence, scope, and credible alternatives.
• Reader effort: task-critical information must be accessible, findable, understandable, and usable; titles, structure, and visuals support that path instead of adding decoration.
• Concrete expression: value-bearing titles, headings, and leads name the actual object and supported result instead of hiding value behind method labels or generic claims.
• Independent review: finished documents must pass four isolated checks for missing context, readability, source reliability, and structural or visual barriers.
• Output fit: the default is self-contained HTML in the source material's dominant language, unless another format or language is requested.
• Delivery boundary: language, format, and publication permission are locked before drafting; failed or missing verification blocks delivery.

Reader's Seat does not publish, overwrite, or send content without explicit authorization. It is intended for substantive documents rather than short-message polishing or unsupported content expansion.

Repository structure, validation, and licenses

Folder	Function
.github/	Runs repository checks on every push and pull request.
agents/	Helps supported agent hosts recognize and present Reader's Seat.
assets/	Provides the self-contained HTML template, local fonts, and chart and diagram runtimes.
config/	Defines document scenarios, module-loading profiles, agent adapters, and the enforceable skill contract.
evals/	Provides frozen trigger, behavior, and cross-agent cases with judging criteria.
examples/	Contains prompts users can adapt to common document tasks.
references/	Provides detailed rules selected for each scenario, format, title, evidence, and visual need.
scripts/	Validates the skill, selects modules, creates HTML report scaffolds, and runs evaluations.
tests/	Verifies language, format, authorization, source integrity, and delivery-receipt enforcement.

Validate an installation:

python3 /path/to/reader-seat/scripts/validate_skill.py

Update an installation:

git -C /path/to/reader-seat pull

Third-party font and visualization licenses are listed in Third-Party Notices.