Plain English Writing
Translates technical content into plain English for non-technical stakeholders by converting jargon into business language, surfacing decisions and impact early, and producing actionable recommendations with clear ownership and timeline. Use when the user asks to simplify technical content, write an executive summary, explain something in layman's terms, rewrite for non-technical audiences, or translate jargon for management or stakeholders. Applies to document types such as technical reports, architecture decision records, incident summaries, risk assessments, and status updates. Distinct from general writing or documentation skills by its focus on audience identification, decision-first structure, and jargon elimination for non-technical readers.
npx skills add github:pantheon-org/tekhne --skill documentation--plain-englishTranslates technical content into decision-ready communication for non-technical stakeholders.
When to Use
Apply when writing for: executives, business managers, compliance/legal, or cross-functional stakeholders without the requesting team’s domain knowledge.
When Not to Use
Skip when writing for engineers or technical specialists. You may optionally apply selective translation for mixed audiences; consider keeping technical depth for peer reviews or developer-facing docs.
Mindset
Technical writing optimises for completeness. Plain-English writing optimises for decisions.
Before writing a single word, ask:
“What does this person need to decide or do — and what is the minimum information required for that?”
The single most reliable fix: move the recommendation to sentence one. Everything else is secondary.
Decision Framework
┌─────────────────────────────────────────────┐
│ OPENING (required) │
│ Problem + Business impact + Action needed │
├─────────────────────────────────────────────┤
│ SUPPORTING CONTEXT (only if needed) │
│ Background / Options / Tradeoffs / Timeline │
├─────────────────────────────────────────────┤
│ APPENDIX (optional) │
│ Technical detail / Metrics / Implementation │
└─────────────────────────────────────────────┘
If a reader only reads the opening, they must still know what to do.
Workflow
Step 1: Identify Audience
executive (zero jargon) | manager (minimal jargon) | compliance/legal (tech translated) | cross-functional (inline definitions). See references/constraints-and-fallbacks.md for the audience depth guide.
If unknown: Audience: unknown. Applying manager-level clarity (fallback).
Step 2: Define the Outcome
One sentence — what must the reader decide or do? Write it before drafting.
Step 3: Draft Key Message First
Opening must contain: problem + business impact + required action.
BAD: [3 paragraphs of background] "...we recommend approving the budget."
GOOD: "We need your approval for a $50k security fix by Thursday. [Impact follows.]"
Step 4: Add Supporting Context
Every action item MUST follow:
[Owner] must [specific action] by [concrete deadline].
BAD: "The database should be optimized."
GOOD: "Database team (Alex) must optimize query performance by March 15."
For acronym-heavy documents, extract all acronyms first:
grep -oE '\b[A-Z]{2,}\b' file.md | sort -u
Define each, then draft — rewrite still leads with key message, not the glossary.
Step 5: Verify
- [ ] Every acronym defined on first use.
- [ ] Opening paragraph contains the key decision or recommendation.
- [ ] All technical terms translated — no unexplained jargon.
- [ ] Every action: [Owner] must [action] by [deadline].
- [ ] Paragraphs are concise and scannable.
Constraints
Hard rules: ALWAYS state audience · ALWAYS define acronyms · ALWAYS lead with key message for executives · NEVER hide decisions · NEVER use passive voice for owned actions · NEVER leave deadlines undefined · NEVER present options without a recommendation.
Fallback paths (state explicitly at the top of your output before writing):
| Situation | Prefix |
|---|---|
| Unknown audience | Audience: unknown. Applying manager-level clarity (fallback). |
| Unknown terminology | Define inline: "TLS (the encryption protocol that keeps data private)" |
| Conflicting priorities | Multiple critical issues — leading with risk and deadlines. |
Full formats: references/constraints-and-fallbacks.md.
Anti-Patterns
Full catalogue: references/anti-patterns.md.
| # | Anti-pattern | Fix |
|---|---|---|
| AP-1 | Jargon without translation | ALWAYS translate or remove every specialist term |
| AP-2 | Buried key decision | ALWAYS put recommendation in sentence one |
| AP-3 | Passive action language | NEVER use passive voice; use [Owner] must [action] by [deadline] |
| AP-4 | Assumed technical context | NEVER assume prior knowledge; explain from first principles |
| AP-5 | Wall of acronyms | ALWAYS define each acronym on first use |
| AP-6 | Recommendations without deadlines | NEVER leave deadlines undefined; undated = ignored |
| AP-7 | Options without a recommendation | NEVER present options without stating which you recommend |
| AP-8 | Skipping audience identification | ALWAYS start with audience and goal |
References
| File | Purpose |
|---|---|
| references/audience-types.md | Audience format and depth guidance |
| references/jargon-translations.md | Technical-to-plain-language mappings |
| references/anti-patterns.md | Full anti-pattern catalogue with BAD/GOOD examples |
| references/before-after-examples.md | Complete before/after rewrites for 5 document types |
| references/constraints-and-fallbacks.md | Full constraint rules, fallback formats, audience depth guide |
External: PlainLanguage.gov
Explicit NEVER/WHY/consequence triplets with side-by-side bad/good examples. Generic warnings score low — concrete failure modes with real impact (security, correctness, performance) score high.
Runtime validation via eval scenarios proving the skill actually changes agent behaviour. Requires evals/ directory, instructions.json, ≥3 scenarios, and ≥80% instruction coverage in summary.json.
Appropriate rigidity for the skill type: strict rules for critical/security skills, balanced steps for process skills, flexible options for tool skills. Mismatched calibration (too rigid or too loose) reduces score.
Expert-only knowledge the AI couldn't derive on its own — domain-specific patterns, production gotchas, decision frameworks, and non-obvious tool behaviour. High scores mean <5% redundancy with what the model already knows.
A clear philosophy plus numbered step-by-step workflows with explicit entry/exit points. Should answer: what is the core principle, how to execute it, and when (and when not) to apply it.
Rich activation keywords and trigger scenarios in the description so the skill fires when needed. Best descriptions read like an exhaustive list of the tasks and concepts that should invoke this skill.
Copy-paste-ready code, complete runnable examples, real file paths and commands. Pseudocode and abstract descriptions score low — concrete, executable content scores high.
SKILL.md as a <100-line navigation hub pointing to focused references/ files. Content should be lazy-loaded on demand, not front-loaded. Includes actionable 'When to Use' conditions per reference.
Proper frontmatter, single-task focus, activation keywords, and cross-harness portability. Skills must work across 40+ agentic harnesses without hardcoded paths or agent-specific references.
Expert-only knowledge the AI couldn't derive on its own — domain-specific patterns, production gotchas, decision frameworks, and non-obvious tool behaviour. High scores mean <5% redundancy with what the model already knows.
A clear philosophy plus numbered step-by-step workflows with explicit entry/exit points. Should answer: what is the core principle, how to execute it, and when (and when not) to apply it.
Explicit NEVER/WHY/consequence triplets with side-by-side bad/good examples. Generic warnings score low — concrete failure modes with real impact (security, correctness, performance) score high.
Proper frontmatter, single-task focus, activation keywords, and cross-harness portability. Skills must work across 40+ agentic harnesses without hardcoded paths or agent-specific references.
SKILL.md as a <100-line navigation hub pointing to focused references/ files. Content should be lazy-loaded on demand, not front-loaded. Includes actionable 'When to Use' conditions per reference.
Appropriate rigidity for the skill type: strict rules for critical/security skills, balanced steps for process skills, flexible options for tool skills. Mismatched calibration (too rigid or too loose) reduces score.
Rich activation keywords and trigger scenarios in the description so the skill fires when needed. Best descriptions read like an exhaustive list of the tasks and concepts that should invoke this skill.
Copy-paste-ready code, complete runnable examples, real file paths and commands. Pseudocode and abstract descriptions score low — concrete, executable content scores high.
Runtime validation via eval scenarios proving the skill actually changes agent behaviour. Requires evals/ directory, instructions.json, ≥3 scenarios, and ≥80% instruction coverage in summary.json.
Expert-only knowledge the AI couldn't derive on its own — domain-specific patterns, production gotchas, decision frameworks, and non-obvious tool behaviour. High scores mean <5% redundancy with what the model already knows.
A clear philosophy plus numbered step-by-step workflows with explicit entry/exit points. Should answer: what is the core principle, how to execute it, and when (and when not) to apply it.
Explicit NEVER/WHY/consequence triplets with side-by-side bad/good examples. Generic warnings score low — concrete failure modes with real impact (security, correctness, performance) score high.
Proper frontmatter, single-task focus, activation keywords, and cross-harness portability. Skills must work across 40+ agentic harnesses without hardcoded paths or agent-specific references.
SKILL.md as a <100-line navigation hub pointing to focused references/ files. Content should be lazy-loaded on demand, not front-loaded. Includes actionable 'When to Use' conditions per reference.
Appropriate rigidity for the skill type: strict rules for critical/security skills, balanced steps for process skills, flexible options for tool skills. Mismatched calibration (too rigid or too loose) reduces score.
Rich activation keywords and trigger scenarios in the description so the skill fires when needed. Best descriptions read like an exhaustive list of the tasks and concepts that should invoke this skill.
Copy-paste-ready code, complete runnable examples, real file paths and commands. Pseudocode and abstract descriptions score low — concrete, executable content scores high.
No eval scenarios available.
Anti-Patterns Reference
Common mistakes when translating technical content to plain English, with examples and fixes.
AP-1: Jargon Without Translation
WHY: Non-technical readers cannot act on unexplained specialist terms. Consequence: Decision delay, confusion, and loss of trust.
| BAD | GOOD |
|---|---|
| “The API returns 429 due to upstream throttling.” | “The system is receiving too many requests; retry windows are being enforced.” |
| “Refactor the monolith into microservices.” | “Break the large application into smaller independent services to reduce release risk.” |
| “The WAF and IDS flagged anomalous egress.” | “The firewall and intrusion detection system flagged unusual outbound traffic.” |
AP-2: Assumed Technical Context
WHY: Stakeholders have varying domain familiarity across roles. Consequence: Gaps in understanding lead to misaligned actions.
| BAD | GOOD |
|---|---|
| “Use RBAC to restrict access.” | “Restrict access so each person can only see data their role requires.” |
| “The p99 latency is 820ms.” | “Response time for 99% of requests is 820ms — well above our 200ms target.” |
| “Scale horizontally with read replicas.” | “Add additional database servers that handle read-only requests to reduce load on the primary database.” |
AP-3: Buried Key Decision
WHY: Time-constrained readers scan only the first section. Consequence: Critical actions are missed or deferred.
| BAD | GOOD |
|---|---|
| Three paragraphs of background, then: “Therefore, we recommend approving the budget.” | Opening paragraph: “We need your approval for a $50k security fix by Thursday.” |
| Risk assessment that lists options without stating a recommendation. | First line: “We recommend Option A. Details and tradeoffs follow.” |
AP-4: Passive Action Language
WHY: Passive voice obscures who is accountable for what. Consequence: Work stalls because nobody owns it.
| BAD | GOOD |
|---|---|
| “The database should be optimized.” | “The database team (Alex Chen) must optimize query performance by Q2.” |
| “It is recommended that penetration testing be conducted.” | “The security team must complete penetration testing before the March 15 release.” |
| “Monitoring should be improved.” | “DevOps (Sam Park) must configure alerting thresholds by end of sprint.” |
AP-5: Skipping Audience Identification
WHY: Tone, depth, and vocabulary must match reader intent and expertise. Consequence: Over-technical or oversimplified content that fails to land.
| BAD | GOOD |
|---|---|
| Start writing immediately without stating who will read it. | “Audience: CFO. Goal: approve Q3 infrastructure budget. Depth: business impact only.” |
AP-6: Wall of Acronyms
WHY: Stacked undefined acronyms create a decoding burden before the reader reaches the message. Consequence: Readers disengage or misread the document entirely.
| BAD | GOOD |
|---|---|
| “The CISO confirmed the WAF, IDS, and SOC 2 audit cover PCI-DSS scope.” | “The Chief Security Officer confirmed our firewall, intrusion detection, and annual security audit cover our payment card data requirements. (Acronyms: CISO, WAF, IDS, SOC 2, PCI-DSS — defined in glossary.)” |
AP-7: Recommendations Without Deadlines
WHY: Undated recommendations are treated as optional. Consequence: No action is taken and risk persists.
| BAD | GOOD |
|---|---|
| “We recommend patching the vulnerability.” | “The security team must deploy the patch by Friday 17:00 to close the exposure window before the weekend.” |
| “Consider migrating to encrypted storage.” | “Engineering must migrate to encrypted storage before the Q3 compliance audit (deadline: June 1).” |
AP-8: Conflating Options with Recommendations
WHY: Presenting options without a recommendation forces the reader to make a technical judgment they lack expertise for. Consequence: Decision escalates unnecessarily or is indefinitely deferred.
| BAD | GOOD |
|---|---|
| “Option A: patch in place. Option B: full redeploy. Option C: defer.” | “We recommend Option A (patch in place). It is fastest and lowest risk. Options B and C are available but each add 2+ days of delay.” |
Audience Types
Use this reference to adapt message structure by reader profile.
Executive Audience
- Decision-focused, short attention window.
- Prefer recommendation first, impact second, risk third.
- Keep to one page when possible.
Suggested structure:
- Recommendation
- Business impact
- Risks and mitigations
- Decision needed and deadline
Manager Audience
- Needs plan, ownership, and timeline.
- Expects tradeoffs and resource implications.
Suggested structure:
- Objective
- Current state and impact
- Options/tradeoffs
- Action plan with owners
Compliance/Legal Audience
- Needs obligations, evidence, and exposure.
- Prefers precise language and traceable facts.
Suggested structure:
- Requirement or risk statement
- Evidence/status
- Gap and remediation steps
- Reporting timeline
Cross-Functional Audience
- Needs context across teams and dependencies.
- Prefers concise shared vocabulary and explicit interfaces.
Suggested structure:
- Summary of change
- Team impacts
- Dependencies and blockers
- Actions by team
Before and After Examples
Concrete rewrites showing the plain-english workflow in practice.
Example 1: Incident Summary (Executive)
Before (engineer-facing):
At 14:32 UTC, a misconfigured Nginx reverse proxy caused a cascading failure in the Kubernetes pod autoscaler, resulting in a 503 surge. MTTR was 47 minutes. Remediation required patching the ConfigMap and rolling restart of the ingress controller. To prevent recurrence, we recommend implementing admission webhooks and enforcing IaC policy gates via OPA Gatekeeper.
After (CEO/CFO-facing):
Decision needed: approve $30k for automated configuration safety checks.
On Tuesday afternoon, an incorrect configuration in our web traffic routing layer caused the payment service to become unavailable for 47 minutes. No customer data was lost. The immediate issue is fixed.
To prevent recurrence, the infrastructure team recommends automated configuration validation — estimated at $30k and two weeks of engineering time. Without this, a similar outage could recur in any future release.
Action required: Engineering VP (Lisa Torres) must approve the $30k budget by Friday 17:00 to begin work in the Q3 sprint.
Example 2: Risk Item (Cross-functional)
Before:
Token replay attacks should be mitigated.
After:
Security team (Alex Chen) must implement session expiry controls by 15 March to prevent unauthorized reuse of login tokens. Without this fix, an attacker who captures a user’s session could continue accessing the account even after the user logs out.
Example 3: Status Update (Unknown Audience — Manager Fallback)
Before (engineering notes):
Profiled query execution plans; found 3 N+1 queries. Added composite index on (user_id, created_at); p99 latency down from 820ms to 190ms. Blocked on DBA approval. Next: profile checkout flow, add read replicas if needed.
After (manager-level):
Summary: Checkout is 4× faster in testing. Production deployment is blocked on database administrator sign-off.
Progress: The team identified and fixed three inefficient database queries. Response time in the test environment dropped from 820ms to 190ms — well within our 200ms target.
Blocker: Database team (Sarah Lin) must approve the change for production by Wed Mar 15 17:00 to meet the Q2 performance milestone.
Next step: Once approved, the team will profile the checkout payment flow and assess whether additional database capacity is needed.
Example 4: Architecture Decision (Compliance/Legal)
Before (technical ADR excerpt):
ADR-042: Migrate from on-prem PostgreSQL to AWS RDS Aurora (Multi-AZ). RPO reduced from ~24h to ~5 minutes (continuous WAL archiving). RTO reduced from ~4h to ~15 minutes. Requires SOC 2 Type II re-assessment for data residency.
After (compliance manager-facing):
Decision needed: approve or reject cloud database migration by [date].
The engineering team proposes moving the customer database from our own data centre to a managed cloud service. This change would reduce the maximum data loss in a disaster from 24 hours to 5 minutes, and reduce recovery time from 4 hours to 15 minutes.
Compliance impact: This change requires a new security certification assessment (SOC 2 Type II) due to a change in where customer data is stored. The assessment takes approximately 3 months and costs an estimated $15k.
Cost increase: $1,200/month.
Planned downtime: 6-hour window during Q3 scheduled maintenance.
Example 5: Security Risk with Conflicting Priorities
Before (two separate tickets):
TICKET-1: EU data breach detected. PII potentially exposed. GDPR notification required. TICKET-2: CVE-2024-XXXX in core dependency. RCE possible. Patch available.
After (executive briefing — conflicting priorities fallback: lead with risk + deadline):
Immediate action required on two security issues. Both deadlines are within 72 hours.
Issue 1 — Customer Data Exposure (Deadline: 54 hours) We detected unauthorized access to our EU customer database 18 hours ago. Some customer personal information (names and email addresses) may have been viewed. We are legally required to notify affected customers and EU regulators within 72 hours of detection (54 hours remain). Legal team (Maria Kovacs) is ready to send notifications pending your approval.
Issue 2 — Security Vulnerability in Core Software (Deadline: Before weekend) A critical security flaw was found in a software library we rely on. If exploited, an attacker could gain unauthorized control of our servers. A fix is available and requires a 4-hour scheduled outage. Engineering (Dan Wu) recommends deploying this Saturday night.
Actions needed from you:
- Legal team (Maria Kovacs) must receive your approval to send EU customer notifications by [specific deadline — 54 hours from detection].
- Engineering (Dan Wu) must receive your approval for the Saturday maintenance window by Friday 17:00.
Constraints and Fallback Paths
Hard Rules
- ALWAYS state the audience explicitly before writing.
- ALWAYS define every acronym on first use — scan the full document after drafting.
- ALWAYS lead with key message (problem + impact + action) for executive audiences.
- NEVER hide decisions after background context; recommendation goes in paragraph one.
- NEVER use passive voice for required actions when ownership is known — name the owner.
- NEVER leave a deadline undefined for a required action; undated recommendations are ignored.
Fallback Paths
Apply when information is missing. State the fallback rule explicitly at the top of your output before writing the document.
| Situation | Rule | Required output prefix |
|---|---|---|
| Unknown audience | Default to manager-level clarity | Audience: unknown. Applying manager-level clarity (fallback). |
| Unknown terminology | Add inline glossary | Unknown terminology detected. Adding inline definitions on first use. |
| Conflicting priorities | Lead with risk + deadline for each | Multiple critical issues — leading with risk and deadlines. |
Unknown Audience
Write Audience: unknown. Applying manager-level clarity (fallback). at the top of your output, then:
- Avoid jargon and translate all metrics to business terms.
- Keep depth moderate — more than a one-liner, less than a technical spec.
- Include owner and deadline for every action item.
Unknown Terminology
When a term’s plain-language equivalent is unclear:
- Write the original term.
- Add a parenthetical definition immediately after: “…using TLS (Transport Layer Security — the encryption protocol that keeps data private in transit)…”
- If many terms are unknown, append a glossary at the end of the document.
Conflicting Priorities
When two or more critical issues compete, open with:
IMMEDIATE ACTION REQUIRED — [N] issues, all deadlines within [timeframe].
Issue 1 — [Plain-language label] (Deadline: [specific time])
[One sentence: what happened, who is affected, what decision is needed.]
Issue 2 — [Plain-language label] (Deadline: [specific time])
[One sentence: what happened, who is affected, what decision is needed.]
Actions needed:
- [Owner] must [action] by [deadline].
- [Owner] must [action] by [deadline].
Do not merge the issues. Each must be independently understandable.
Action Item Format Rule
Every action item — in any document type — MUST follow this pattern:
[Team or person name] must [specific action] by [concrete date or deadline].
If the owner is unknown, write OWNER TBD — never leave an action ownerless or undated.
Examples:
| BAD | GOOD |
|---|---|
| “The database should be optimized.” | “Database team (Alex Chen) must optimize query performance by March 15.” |
| “Monitoring needs to be improved.” | “DevOps (Sam Park) must configure alerting thresholds before the Q2 release (May 1).” |
| “Legal review is recommended.” | “Legal team (Maria Kovacs) must complete contract review by Friday 17:00.” |
Audience Depth Guide
| Audience | Jargon tolerance | Decision format | Metric translation |
|---|---|---|---|
| Executive (CEO/CFO/board) | Zero — all terms translated | Single sentence + approve/reject | Business impact only (cost, time, risk) |
| Manager/VP | Minimal — common business terms OK | Recommendation + options | Outcomes (e.g., “4× faster”) |
| Compliance/legal | Domain terms OK; tech terms translated | Risk + regulatory impact + action | Deadlines and penalties |
| Cross-functional | Light jargon with parenthetical definitions | Problem + impact + owner | Mixed — follow audience mix |
Jargon Translations
Use these mappings to keep technical meaning while improving readability.
| Technical Term | Plain English |
|---|---|
| API | connection point between systems |
| Database | central data storage |
| Microservice | small independent application component |
| Container | packaged runtime unit |
| Latency | response delay |
| Throughput | processing capacity over time |
| Refactor | improve internal structure without changing outcome |
| Technical debt | deferred maintenance work |
| Incident response | organized process to handle security/service events |
| Dependency upgrade | update a supporting software component |