Inspect a skill

Markdown editor and agent skill inspector

A quiet Markdown editor that stays in your browser.
1---
2title: Better meetings
3subtitle: A short guide, written and previewed in mdprev
4tags: [meetings, teamwork, notes]
5updated: 2026-11-15
6---
7
8# Better meetings
9
10Most meetings run long because nobody decided what they were for. Write the goal as one sentence, then invite only the people who can act on it. A chain of ~~reply-all~~ emails is not a meeting.
11
12> [!TIP]
13> Edit the text and the page follows. To turn this document into slides, choose **Present** (or press Esc, then **S**).
14
15## Before
16
17- [x] Write the goal as one sentence
18- [x] Invite the people who decide
19- [ ] Send the agenda a day early
20- [ ] Book fifteen minutes, not an hour
21
22## The agenda
23
24| Slot | Minutes | What happens |
25| :--------- | ------: | :------------------------------ |
26| Goal | 1 | Read the one-sentence goal |
27| Updates | 3 | Only what changed |
28| Decision | 4 | Choose, or choose who chooses |
29| Next steps | 2 | Name an owner and a date |
30
31## Why it matters
32
33A meeting costs the time of everyone in it:[^1]
34
35$$
36\text{cost} = n \times t \times r
37$$
38
39where $n$ people spend $t$ hours at an hourly rate $r$.
40
41## When to skip it
42
43```mermaid
44flowchart LR
45 A[Need a decision?] -->|Yes| B[Hold the meeting]
46 A -->|No| C[Send a note]
47```
48
49## A template
50
51```md
52# Weekly sync
53
54Goal: one sentence.
55Decision: what we chose.
56Next: owner, date.
57```

Better meetings

Most meetings run long because nobody decided what they were for. Write the goal as one sentence, then invite only the people who can act on it. A chain of reply-all emails is not a meeting.

Tip

Edit the text and the page follows. To turn this document into slides, choose Present (or press Esc, then S).

Before

  • Write the goal as one sentence
  • Invite the people who decide
  • Send the agenda a day early
  • Book fifteen minutes, not an hour

The agenda

Slot Minutes What happens
Goal 1 Read the one-sentence goal
Updates 3 Only what changed
Decision 4 Choose, or choose who chooses
Next steps 2 Name an owner and a date

Why it matters

A meeting costs the time of everyone in it:

cost=n×t×r\text{cost} = n \times t \times r

where nn people spend tt hours at an hourly rate rr.

When to skip it

MERMAID
flowchart LR
  A[Need a decision?] -->|Yes| B[Hold the meeting]
  A -->|No| C[Send a note]

A template

MD
# Weekly sync

Goal: one sentence.
Decision: what we chose.
Next: owner, date.
What is mdprev? Free, runs in your browser, also checks agent skills
Markdown editor

A quiet editor for everything else.

Split preview, a serif reading view, and exports that look like you meant them to.

  • PreviewGitHub-flavoured Markdown, tables, callouts, footnotes, KaTeX and Mermaid, rendered off the main thread.
  • ExportMarkdown, standalone HTML, PDF with contents, Word and slides (one page per slide).
  • ShareA link that carries the document itself. No account and no copy on a server.
  • OfflineInstall it as an app (Ctrl+/ then “Install mdprev as an app”, or your browser’s Install or Add to Home Screen). Editing and preview keep working without a network.
  • SlidesPresent any document, with a presenter window for notes and a timer.
  • ZenA quiet writing column with typewriter scroll and focus by paragraph or sentence.
Skill inspector

See what a skill is before it is in your agent.

Paste a GitHub link or open a zip. You get a plain report, with no scores and no badges. Then edit a copy in the browser, see the diff against the original and download it as a zip.

Does it follow the spec?

Checks the frontmatter against the Agent Skills specification and tells you where the folder name does not match.

What does it cost?

Estimates tokens at three levels: always loaded, loaded when the skill triggers, and every file. Open a whole collection to see the total.

What would it run?

Lists scripts and flags patterns worth reading first, such as shell calls and recursive deletes. A static scan, not a safety guarantee. Nothing is run.

Inspect a skill · How the inspector works

Privacy

Your files stay in your browser.

Nothing you open is uploaded, and there is no analytics script on these pages. Reading a skill from GitHub fetches it straight from GitHub to your browser. What is stored, and where.

  • No upload
  • No account
  • No analytics
  • Works offline
Questions

Common questions

What is an agent skill?

A folder with a SKILL.md file (instructions for an AI agent) and optional scripts and reference files. Agents read the short description of every skill at startup and load the rest only when a task needs it.

Can mdprev tell me a skill is safe?

No. It points at things worth reading, using fixed rules. A clean report does not mean a script is safe, and a flag does not mean it is harmful. Read what it flags.

Are the token numbers exact?

No. They are estimates from character count (about four characters per token). Real counts depend on the model.

Do I need an account?

No. There is nothing to sign in to, and nothing you open is uploaded.