reader-comprehension-pass
Skill2 starsChanged 6 months agoMITView on GitHub
Make technical writing land with its reader — add figures for hard-to-grasp concepts, define jargon in plain terms at first mention, and audit the draft for unintuitive spots before delivering. Use when writing or revising research documents, manuscripts, status notes, handovers, or explanatory artifacts, and before handing any such draft to the user.
In this repository10 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Keep the distance between a concept's first mention and its explanation minimal — order sections so the reader has the needed background at …
- Distinctive: line 29
Ask where the text supplied each prerequisite, rather than letting an expert fill gaps from outside knowledge.
- Distinctive: line 33
Use an independent reader when available and authorized.
- Distinctive: line 37
When translating slides into a paper, check that spoken bridges now exist in prose, algorithm and figure identities still match, and …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reader comprehension pass
The user reads agent-authored technical documents as a researcher would, and repeatedly hits concepts that prose alone does not convey. Apply these rules while drafting, and run the audit before delivering.
Illustrate hard concepts with figures
- When a central concept is spatial, geometric, or quantitative and the prose is doing heavy lifting, draw it instead of explaining harder. Prefer a figure carrying real values from the work over an abstract sketch — e.g. "search box floor" only became legible as a diagram showing an actual proposal point (8.15, 4.06) clipped to the constraint-box wall at (6.8, 5.2).
Define jargon at first mention
- Give every domain term the target reader may not know a plain-terms definition at its first use (the user could not parse "Sn-based non-convergence" or "censored loss" left bare in a manuscript).
- Keep the distance between a concept's first mention and its explanation minimal — order sections so the reader has the needed background at the point of first mention, never referencing a concept long before it is explained.
Audit before delivering
- Before handing over a draft, reread it as a first-time reader and list every term or concept that would be unintuitive. Fix each one (figure, definition, reordering) or flag it explicitly in your response. The user expects this proactively — do not wait to be asked "what else is not intuitive?"
Test an unprompted explanation
For a substantial explanation or paper revision, ask a fresh reader to retell the problem, proposed insight, essential operation, and demonstrated finding using only the audience-facing material. Give the audience brief but withhold author notes, later slides in a prefix test, expected answers, and prior critiques. Ask where the text supplied each prerequisite, rather than letting an expert fill gaps from outside knowledge. Record what was misunderstood or could not be answered, and repair the first causal gap before adding more definitions.
Use an independent reader when available and authorized; otherwise label the pass self-review. A model reader is a proxy, not proof of human comprehension or recall. A small wording edit needs a proportionate local check, not a new team.
When translating slides into a paper, check that spoken bridges now exist in prose, algorithm and figure identities still match, and compression has retained assumptions. Do not demand a fixed sentence count or a rhetorical question. Plain language must preserve the original mathematical and empirical claim.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

