semantic-markdown-writer
Skill3 starsChanged 5 days agoMITView on GitHub
Create and edit Semantic Markdown (.smd) documents (PRDs, ADRs, RFCs/design docs, runbooks, API references, status reports, meeting notes, postmortems, release notes, OKRs, onboarding guides, test plans, PR descriptions, specs and plans) with templates and a validator that fixes mistakes. Use it whenever the user asks to write, draft, create, convert, restructure or update a .smd file, asks for a spec/PRD/ADR/runbook/status report "in smd" or "semantic markdown", wants a Markdown doc converted to .smd, or wants a document that both people and AI agents will read.
1 other project uses a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- bislink360/styled-markdownedited, first seen Nov 8, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 1
- bislink360/styled-markdown★ 394% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- smd-releaseSkill
- sonarqube-scanSkill
- version-controlSkill
- semantic-markdown-readerSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Start from a template: when one fits: smd init docs/name.smd --template prd --title "Saved searches" (today's date filled in.
- Very specific: line 35
Fill it with real content: Delete sections that don't apply.
- Very specific: line 62
Meaning over decoration: Use :::warning, :::danger or :::decision for things that matter.
≈ “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
Writing Semantic Markdown (.smd)
.smd is Markdown plus a small, validated vocabulary: callouts, decisions, risks, API blocks, tasks with owners and due dates, KPIs, diagrams, and audience blocks. People read it in a styled preview, AI agents through a compact "agent view". Write for both.
Tool (Node 22+, no install): node <this-skill-dir>/scripts/smd.cjs <command> (or smd if it's on PATH).
Workflow
Start from a template when one fits:
smd init docs/name.smd --template prd --title "Saved searches"(today's date filled in; raw files inassets/templates/).Request Template Product requirements, feature spec prdArchitecture/technical decision adrDesign proposal, RFC, tech spec rfcOn-call / operational procedure runbookEndpoint reference apiWeekly/monthly update status-reportMeeting summary with actions meeting-notesIncident review postmortemRelease announcement, upgrade guide release-notesObjectives and key results okrsNew team member guide onboardingTest scope, strategy and cases test-planPull request description pr-descriptionSpec-driven work for an agent: spec → plan → tasks spec,plan,tasksOtherwise start from front matter + headings. For a status report on work tracked in
.smddocs, draft it withsmd report docs/ --since 2026-09-01 -o docs/status.smd --title "Checkout squad"(or--since <git-ref>): tasks done and added, open and overdue tasks, decisions and high-impact risks, each linked. Then write the summary and cut what doesn't matter.- Fill it with real content. Delete sections that don't apply; never leave placeholders such as
—,@ownerorYYYY-MM-DD. If you don't know a value, ask, or leave a:::questionthat says what's missing. - Format, validate and fix:
smd fmt docs/name.smd, thensmd validate docs/name.smd --fix.fmtchanges only layout.--fixrepairs what has one clear repair (typos in names and values, colors,yes/nobooleans, date formats, unclosed:::or fences at the end). Fix the rest yourself from each error's line:column, message and rule code. A document is done only at 0 errors. - Check what agents will see:
smd agent docs/name.smd --brief. If it's long, move narrative into{agent=skip}sections or:::humanblocks.
Spend tokens on meaning
Output tokens cost about 5× input tokens. Write meaning: decisions, risks, tasks, :::agent rules, {agent=skip} on narrative. Leave colors, badges and layout (tabs, columns, cards) to people unless asked. The syntax reference is split by topic: read references/syntax.md (a short index), then only the topic file for the construct you're unsure of, and let smd validate --fix catch the rest.
Rules you must follow
- Front matter first:
smd: 2(new documents; keep an existing document's value),title, a precise one- or two-sentencesummary,status(draft·review·approved·deprecated·archived),owners,updated(YYYY-MM-DD). Don't repeat the title as a# H1; start at##. Another language addslang:(de,es,fr,ja,pt,zh, or a tag likept-BR). - Blocks are
:::name{attrs} Title…:::. Attributes go directly after the name, no space. A bare:::closes the innermost block. Write outer containers with more colons (::::tabs). - Only known names. Blocks:
noteinfotipsuccesswarningdangerquestiondetailscardboxtabs/tabcolumns/columnstepstimelinefigureglossaryquotedecisionriskrisk-matrixchangelogapiagenthumaninclude. Inline::badge:status:priority:due:metric:progress:kbd:mention:ref. - Values exactly as specified: decision
statusproposed · accepted · rejected · superseded · deprecated; riskimpact/likelihoodlow · medium · high · critical; apimethodGET, POST… and a requiredpath; priority P0–P4; datesYYYY-MM-DD. Figures, glossary, quotes and changelogs have their own forms (syntax/figures-and-notes.md,syntax/project.md). - Named colors only (red orange amber yellow green teal cyan blue indigo purple pink gray muted accent) unless a brand hex is required.
- Tasks:
- [ ] Verb-first task :priority[P1] @owner :due[2026-10-15], one owner where possible. - Code: give fences a language; embed real source with
file="../src/x.ts" lines="10-24"(empty body) instead of pasting it. Shared text goes in one file, pulled in with:::include(syntax/developer.md). - Diagrams:
```mermaidwith a valid first line (flowchart LR,sequenceDiagram,gantt, …). - Facts that repeat (version, product name, date) go in the front matter once and appear as
{{version}}. Quote versions (version: "2.10";syntax/inline.md). - Spec v2 shows text as written: straight quotes stay straight (type
—or addtypography: true), onlyhttps://…,www.…and e-mail addresses become links, any:name[…]is a directive (\:name[…]is text). Don't change ansmd: 1document's value while editing;smd migrate FILE --to 2moves it (syntax/spec-v2.md).
Writing for both audiences
references/style-guide.md covers this in depth. In short:
- Meaning over decoration. Use
:::warning,:::dangeror:::decisionfor things that matter. Color must never be the only signal. - Hard constraints for implementers go in one
:::agentblock near the end. Agents always receive it, even when they read one section. - Narrative, history and research go under
## Background {agent=skip}or in:::human. People still see it; agents skip it. - Unresolved decisions go in
:::question, with who decides and by when; the interim fallback goes in the:::agentblock. - Headings name the content ("Requirements", "API", "Rollout") because agents select sections by heading.
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.

