Install
$ agentstack add skill-nikdumroese-plain-language-skill-plain-language-iso-24495 ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Plain language (ISO 24495-1:2023)
Apply the four principles to every document: Relevant, Findable, Understandable, Usable. Readers must be able to find what they need, understand it, and use it.
The workflow below is a loop, not a checklist. If new reader information appears at step 3, return to step 1.
References
Load on demand:
references/principles.md— full principles and guidelines.references/checklist.md— 42-item pre-flight review.references/reader-test.md— reader evaluation. External docs only; skip for internal memos.assets/template.md— starting shape for the output. Use it as the default structure.
Step 1 — Identify readers and format (Principle 1)
Before drafting, know:
- Readers: roles or teams. If several, name the primary audience.
- Purpose: what they must do after reading (decide, approve, act, learn).
- Prior knowledge: what they know; adjust vocabulary.
- Reading context: phone, desktop, meeting, Slack link.
- Time available.
- Language: confirm before drafting in English if the reader group is not native.
- Format: a video, form, FAQ, or dashboard may serve better than prose.
Ethics: accurate content, no misleading claims, no hidden content the reader needs.
If any of the above is missing, ask one focused question before drafting. Do not skip.
Step 2 — Structure for findability (Principle 2)
- The most important message goes at the top. The reader knows the ask, decision, or answer within the first screen.
- Multiple audiences get their own labelled section.
- Warnings before instructions.
- Processes in chronological order.
- Headings describe what follows. Scanning headings alone reconstructs the argument.
- Supplementary detail at the end, in a labelled section.
- Apply prominence (bold, callouts), proximity (group related items), similarity (same shape for same function). Use white space to separate ideas.
Step 3 — Write for understandability (Principle 3)
- Familiar words. Keep a technical term only if the everyday word is less accurate.
- One idea per sentence. Average 15–20 words.
- One idea per paragraph. Front-load the main point.
- Active voice. Name the actor.
- Direct address. Use "you".
- Direct requests. No performative politeness ("please" every paragraph, "thank you for your time"). No scolding, no condescension.
- Diagrams for spatial, temporal, or relational content.
- Define technical terms on first use or in a glossary.
- Consistent terminology. Same term for the same concept. No synonyms.
Step 4 — Make it usable (Principle 4)
- Self-review with
references/checklist.mdbefore delivering. - Include an explicit response mechanism: an approval memo has "Approved / Approved with comments / Not approved"; an RFC points to a comment thread or decision meeting; a runbook names who to page.
- Do not push reader testing for internal memos. For external documents (public terms, customer policy, regulator submissions), load
references/reader-test.md. - If a reader comes back confused, revise. That is the signal.
Required output shape
Every document must include:
- Headline stating what the document is and what it asks for.
- Header block with author, date (
YYYY-MM-DD), version, status, reading time, and reader groups.assets/template.mdshows the shape. - Main content grouped by reader need, not by author's mental hierarchy.
- Warnings before the instructions they warn about.
- Glossary or inline definitions for technical terms the primary reader may not know.
- Response mechanism stating how the reader should respond or act.
- Related documents section at the end.
Multi-audience routing (recommended, not required)
When a document has more than one reader group, help each reader find their section quickly. ISO 24495 requires that structure supports findability but does not prescribe a specific pattern. Pick one:
- "Are you the right reader?" section listing reader groups and what each is asked to do. Good default for approval memos.
- Routing table at the top: audience → section. Good for reference docs with many audiences.
- Labelled sections by audience with a one-line pointer in the summary. Good for short documents.
Skip for single-audience documents.
Anti-patterns
Do not do the following.
Structure
- Do not start with background or context. Start with the ask.
- Do not bury the decision in a wall of text.
- Do not include meta-content ("This document will explain…").
- Do not use rhetorical questions as headings ("Why do we need this?").
- Do not nest lists more than two levels. Use a table.
Attribution
- Do not personalise sections with names ("For Jane Doe"). Address roles or teams. People move; roles persist. If a POC is stable, put the name in an ownership table, not a heading.
- Do not attribute ownership to multiple teams without naming a primary. "Owned by Data, Analytics, and Engineering" means owned by nobody.
- Do not use author-centric framing ("I want to propose…", "We have been thinking…").
Language
- Do not use synonyms for the same concept.
- Do not use passive voice when active is shorter and clearer.
- Do not use jargon without defining it.
- Do not chain acronyms ("SRE-owned CI/CD via IaC").
- Do not use "TL;DR". Use "Summary" or a plain headline.
- Do not use contractions in documents for non-native English readers.
- Do not use filler ("In today's fast-paced world", "It goes without saying").
- Do not use marketing verbs: leverage, synergise, empower, unlock, drive.
- Do not use weasel words: we believe, arguably, potentially, somewhat, it seems.
- Do not use vague quantifiers: a lot, many, significant, recently, soon. Use numbers and dates.
- Do not mix modal verbs. RFC 2119:
must= requirement,should= strong recommendation,may= optional. Do not treat them as interchangeable.
Content
- Do not include content the primary reader does not need.
- Do not state a constraint without its reason. "We must use PostgreSQL" needs a why.
- Do not leave success criteria vague ("we will monitor and iterate").
- Do not leave review requests vague ("please review and provide feedback").
- Do not compare to strawmen. Compare to real alternatives.
- Do not use false precision. "~$50/month" is fine; "up to 40% faster" without a baseline is not.
- Do not omit dates, version, or author from the header.
- Do not cross-reference other documents without a link or ID.
Tone
- Do not scold, condescend, or lecture.
- Do not use passive-aggressive framing ("as has been repeatedly mentioned…").
- Do not manufacture consensus ("we all agree that…").
After delivering
Offer a revision pass if a reader comes back confused. Do not push reader testing for internal memos.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: nikdumroese
- Source: nikdumroese/plain-language-skill
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.