workplace-violence-prevention

name	workplace-violence-prevention
description	Produces a consultant-grade workplace-violence-prevention (WPV) program for a named healthcare service — the worksite hazard analysis, the type-1-4 violence taxonomy, environmental and administrative controls, and the confidential incident log. Use this skill whenever a user asks to prevent workplace violence or aggression in healthcare, build a WPV program or plan, assess violence risk in an ED / mental-health / care setting, design de-escalation or security controls, or set up violence incident reporting. It leads with environmental and administrative design (controlled access, sightlines, alarms / duress, staffing, lone-working) before reactive measures, classifies violence by source (type 1 criminal -> type 4 personal), and records a de-identified worksite analysis with small-cell suppression. It refuses a program headlined by "personal alarms / self-defence training" alone. Grounded in OSHA 3148 + §5(a)(1) and Cal/OSHA 8 CCR 3342. Decision-support only; review by a competent person.
license	Apache-2.0
metadata	{"author":"eyekyam","version":"1.0","category":"occ-health","tier":2,"audience":["M","C","F"],"industry":["All"],"jurisdiction":["All"],"status":"stable","plugin":"hse-healthcare","hse_reviewed_by":"","hse_reviewed_date":""}

Workplace-Violence Prevention

A consultant-grade, environmental-and-administrative-control-first workplace-violence (WPV) prevention program for a named healthcare service or unit — an emergency department, a mental-health unit, a reception / triage area, a community or lone home-visit service, an ambulance service, or a residential care setting — never a generic "a hospital". Its entire reason to exist is that WPV is controlled by designing out the exposure, not by reactive measures: every program first records the worksite hazard analysis (records review, incident log, walkthrough, staff input) and then leads with environmental / engineering controls (controlled access, sightlines, alarm / duress systems, secure design) and administrative controls (staffing / skill-mix, lone-working procedures, known-risk-patient flagging, no-retaliation reporting), with de-escalation, response training, and personal alarms as the documented residual lines. A bare "issue personal alarms and run self-defence training" is refused as the headline — a reactive measure is the last line, never the primary control where the exposure could be designed out.

It forces the single lever that separates a defensible artifact from copy-paste paperwork: the worksite hazard analysis plus the environmental-and-administrative-before-reactive hierarchy of controls, with the incident never identifying a person. Workplace-violence incident data is special-category health data (PHI) — a named victim, assailant, or known-risk patient and any behavioural-health flag are scrubbed to role labels before drafting, the worksite analysis and incident log in the circulated artifact are de-identified and aggregated (never line-level), and an incident category with fewer than 5 individuals is suppressed (small-cell back-calculation guarded — a 2-incident category on a named ward de-anonymizes the people involved). The skill never emits a re-identification key file — the key is an instruction to the competent person, held separately and access-controlled. Grounded in OSHA Publication 3148 (2016) Guidelines for Preventing Workplace Violence for Healthcare and Social Service Workers (the five program elements — management commitment & worker participation, worksite analysis, hazard prevention & control, training, recordkeeping & evaluation), enforced under the OSH Act §5(a)(1) General Duty Clause where a recognized WPV hazard exists, the enforceable Cal/OSHA Workplace Violence Prevention in Health Care standard, 8 CCR 3342 (written plan, type-1-4 taxonomy, employee involvement, violent-incident log), the UK HSE management of work-related violence guidance with NICE NG10, and India occupational-safety provisions via hse-india. Decision-support only; a competent person (healthcare security / WPV / occupational-health professional) must review the output.

When to use this skill

Use this skill when the user needs a workplace-violence-prevention program for a concrete healthcare service — for example "prevent violence in the emergency department", "build a WPV program for the mental-health unit", "assess aggression risk for the community lone-working team", "design reception / triage security controls", or "set up confidential violence incident reporting". It is not for a generic "how do I stop workplace violence?" answer: the Workflow intake below forces the named service and where / when violence occurs, the violence type(s), and the worksite hazard analysis before any control plan, refuses a vague "a hospital" request, and refuses a reactive-led "issue personal alarms / run self-defence training" treatment where the exposure could be designed out by environmental or administrative controls.

Data Protection & De-identification (MANDATORY — apply before drafting)

Apply this BEFORE you draft anything. Treat injury, illness, and any health detail as the highest sensitivity. Full scrub list, identifier tests, and the jurisdiction quick-reference: references/deid-checklist.md.

1. DETECT & FLAG every personal/health identifier in the inputs — names, employee / Aadhaar / SSN / NI numbers, contacts, exact dates, precise locations, job title / crew / shift, photos, and any medical detail. List what you found before drafting. If unsure whether something is identifying, treat it as identifying.
2. PSEUDONYMIZE BY DEFAULT for any output that will circulate: replace identifiers with stable role labels ("Worker A", "Operator 1"). Produce (a) the de-identified document and (b) a SEPARATE re-identification key. Never put the key or any name↔label mapping in the document. Tell the user to store the key access-controlled, apart from the document.
3. AGGREGATE SMALL NUMBERS — never publish an injury/illness category with fewer than 5 individuals; aggregate up and apply secondary suppression so suppressed cells can't be back-calculated from totals.
4. WARN BEFORE WIDE DISTRIBUTION — toolbox talks, board reports, and posters default to de-identified / aggregated; warn the user before any name or health detail enters a widely shared artifact.
5. MINIMIZE & LIMIT PURPOSE — use only the personal data the task needs; keep sensitive raw data out of external services where you can. When in doubt, ask before including it.

This skill is one of the catalog's highest-PHI artifacts. The reinforced healthcare PHI extension — the named-victim / assailant / known-risk-patient rule, the behavioural-health-flag rule, the <5 small-cell suppression with secondary back-calculation guard, and the re-identification-key-separation instruction — lives in references/deid-checklist.md and the Workflow de-id step below (it is NOT in the byte-identical block above).

Knowledge base (read ONE matching file — never load all)

Resolve the user's jurisdiction first. Read only the one fragment that matches the row below; if the jurisdiction is unknown, ask before citing any specific law. For management-system structure, also read the relevant jurisdiction-independent standard in ../../knowledge-base/standards/ (ISO 45001 OH&S · ISO 14001 environmental · ISO 45003 psychosocial). Always apply ../../knowledge-base/prompt-snippets/hierarchy-of-controls.md (KB-SNIP-HOC) to every control recommendation. For any benchmark/figure, look up the ID in the relevant _registry.yaml, then read ONLY the named file — and quote its source+year.

Jurisdiction / scope	Read
WPV control spine (every run)	../../knowledge-base/prompt-snippets/wpv-controls.md (KB-SNIP-WPV-CONTROLS) — the worksite analysis → environmental/engineering → administrative → training/response-residual WPV control gate: a program led by "de-escalation training / panic buttons" alone, with no environmental or administrative controls, is rejected as reactive-led; incident reporting is small-cell-suppressed
Healthcare clause cross-walk (every run)	../../knowledge-base/prompt-snippets/healthcare-clause-map.md (KB-SNIP-HEALTHCARE-CLAUSE-MAP) — the bundle-shared ISO 45001 6.1.2 + clinical PPE-last cross-walk that keeps the five hse-healthcare skills consistent
USA + California	../../knowledge-base/regulatory/wpv-osha-3148.md (KB-REG-WPV-OSHA3148) — OSHA Publication 3148 (2016) five-element WPV program map (management commitment & worker participation · worksite analysis · hazard prevention & control · training · recordkeeping & evaluation) enforced under the OSH Act §5(a)(1) General Duty Clause where a recognized WPV hazard exists, plus the type-1-4 violence taxonomy and the Cal/OSHA 8 CCR 3342 written-plan / violent-incident-log standard for California healthcare employers (cite the program-element topics + the General Duty Clause, never paste the guideline; cite 8 CCR 3342 as the binding standard in California)
EU / UK	../../knowledge-base/standards/iso-45003.md (KB-STD-ISO45003) — ISO 45003 psychosocial-risk grounding for the aggression/violence aftermath; combine with the UK HSE management of work-related-violence guidance and NICE NG10 (violence & aggression: short-term management in health/community settings) for the binding UK leg (cite the guidance, never paste it)
India	India occupational-safety provisions — defers to hse-india, mandatory state detection (state forms / rules resolved via KB-REG-IN-STATEFORMS); emit [GAP], never a national form number
Unknown	Ask before citing any specific law

This skill always grounds in KB-STD-ISO45001 (6.1.2, with the KB-STD-ISO45003 psychosocial link for the aggression / violence aftermath) and leads every program with the WPV environmental-and-administrative-first spine KB-SNIP-WPV-CONTROLS (worksite analysis → environmental / engineering → administrative → training / response residual), classifying hazards by the type-1-4 taxonomy and grounding the five program elements in KB-REG-WPV-OSHA3148, aligns with the other hse-healthcare skills through KB-SNIP-HEALTHCARE-CLAUSE-MAP, applies KB-SNIP-HOC to every control, and reuses the KB-SNIP-ARCHETYPES subagent roster. The control selection is a structured environmental-and-administrative-control-first method over the named service's worksite analysis + the cited OSHA 3148 / §5(a)(1) / Cal-OSHA framework — not a calculation (no new engine). For a US site ground in KB-REG-WPV-OSHA3148 (cite §5(a)(1); cite 8 CCR 3342 as the binding standard in California); for the UK lean on the HSE work-related-violence guidance + NICE NG10; for India, resolve the state via hse-india (mandatory state detection) and emit a literal [GAP] where a state return is owed — never a minted national form number. The rule-9 manifest is references/_skill-kb.md.

Workflow

Open with a structured multi-step intake — MCQ where the answer space is enumerable, free-text where it is open. Ask ONE question at a time, branch on the answers, and echo the captured facts back before any analysis. Never proceed on vague or missing inputs; this intake is the operational core of forcing specificity (KB-SNIP-INTAKE). (Intake is a Workflow convention, not a sixth block.)

De-identify FIRST (the highest-PHI step — before any drafting). Run the deid block + references/deid-checklist.md BEFORE the intake echo-back drives any analysis. Workplace-violence incident data is special-category health data: scrub any named victim, assailant, or known-risk patient and any behavioural-health flag (role labels only — "Worker A", "the patient"; a behavioural-health flag is held in a separate confidential clinical record, never written into a circulated program), and any patient identifier anywhere in the inputs. Apply <5 small-cell suppression (with secondary suppression) to every incident category — a 2-incident category on a named ward de-anonymizes the people involved — and produce a SEPARATE access-controlled re-identification key — never co-located with the program and never emitted as a key file.

Run the WPV intake one question at a time (full coverage contract + branch map in references/intake.md). Refuse to plan for "a hospital": you need the named service + where / when violence occurs + the violence type(s) before any control plan. Refuse a reactive-led "issue personal alarms / run self-defence training" treatment where the exposure could be designed out by environmental or administrative controls.

1. The named service & exposure (free-text — the specificity anchor) — "Name the exact unit / service (ED, mental-health unit, reception / triage, community / lone home-visit, ambulance, care home) and where / when violence occurs. Refuse 'a hospital' / 'the ward' — the program is service-specific."
2. Violence type(s) (multi-select — the OSHA / NIOSH taxonomy) — type 1 criminal-intent · type 2 customer / client / patient (the dominant healthcare type) · type 3 worker-on-worker · type 4 personal-relationship. Each type drives a different control set — classify before controlling.
3. Worksite hazard analysis (free-text — the defensibility anchor) — "The records / incident review, the walkthrough findings, and the employee survey. Use de-identified, aggregated incident data — never a named victim or assailant. A program with no worksite hazard analysis fails." The analysis is the basis of every control.
4. Environmental & administrative controls (multi-select — asked BEFORE reactive / PPE measures) — controlled access & egress · sightlines / reception design · alarm / duress / panic systems · CCTV · waiting-area design · staffing & skill-mix · lone-working procedures · known-risk-patient flagging (handled per data-protection rules — de-identified). A program that jumps to "personal alarms / restraint training" without addressing environmental design and staffing is FLAGGED as reactive-led and pushed up the hierarchy.
5. De-escalation, response & training (residual) (free-text) — the de-escalation procedure, the response / security protocol, post-incident support, and the training plan — the documented residual lines, never the headline control.
6. Jurisdiction (mcq) — USA (OSHA 3148 + §5(a)(1); Cal/OSHA 8 CCR 3342 where applicable) / UK (HSE work-related-violence + NICE NG10) / India / Other / Unknown. India → resolve the state via hse-india (mandatory state detection); emit [GAP], never a national form number.

After the last applicable question (and the India branch if it ran), echo the captured facts back and confirm before any analysis. Never proceed on a vague or missing input — a missing input is a [GAP], never an invented incident or count.

Then: record the worksite hazard analysis first (de-identified, aggregated; KB-SNIP-WPV-CONTROLS); classify the exposure by the type-1-4 taxonomy; rank the controls up the hierarchy via the controls engine (eliminate / substitute exposure where possible → environmental / engineering: controlled access, sightlines, alarms / duress, secure design → administrative: staffing / skill-mix, lone-working, flagging, procedures → de-escalation / response / training and personal alarms LAST — a program led by reactive measures / personal alarms / "self-defence training" with no environmental or administrative control is a FLAG pushed up the hierarchy); author the de-escalation & response protocol + post-incident support; build the confidential, de-identified / aggregated WPV incident log structure (with <5 small-cell suppression); frame the residual violence risk via risk_matrix; make every action a SMART action via smart_actions → validate the draft against references/QUALITY_CHECKLIST.md → produce the output via the Output format section below. The domain method is in references/METHODOLOGY.md.

Agentic Execution (Orchestration Block)

You are the ORCHESTRATOR for this skill. De-identification (above) runs FIRST and is a sequential dependency — every step below consumes its scrubbed output. Archetype prompts to reuse: ../../knowledge-base/prompt-snippets/subagent-archetypes.md (KB-SNIP-ARCHETYPES).

Step 0 — Triage: fan out at all?

Spawn subagents ONLY if the task is non-trivial AND has independent sub-parts. Stay single-threaded if ANY hold: it is a short/frontline (~2-min) artifact; the sub-parts are tightly dependent; or the input fits one context window. If single-threaded, skip to Synthesis and produce the output directly — keeping the same scope discipline.

Step 1 — Plan

Decompose into INDEPENDENT jobs. Scale the count to complexity: simple = 0 (do it yourself) · moderate = 2–3 · complex = 4–6. Never exceed MAX=6.

Step 2 — Fan out (parallel subagents)

Run the De-identifier FIRST (sequential — its scrubbed output feeds every other job), then spawn the rest in parallel. Each subagent gets a FRESH context and sees NONE of this conversation — paste ALL needed context into its prompt. Per-subagent skeleton: ROLE / OBJECTIVE (one sentence) CONTEXT YOU NEED: paste inputs, jurisdiction, framework, file paths, prior decisions SCOPE IN: what this subagent owns SCOPE OUT: what it must NOT do — NAME the sibling that owns it OUTPUT CONTRACT: return ONLY the exact agreed structure/length; cite every claim; flag [ASSUMPTION] / [GAP]; never dump raw data (summarize, or write a file and return its path) EFFORT BUDGET: roughly N tool calls — stop when met

Step 3 — Synthesis (you)

Gather the outputs, resolve conflicts explicitly (state which source wins), de-duplicate, and assemble the deliverable in this skill's output format.

Step 4 — SME Review & Sign-off (MANDATORY — regulatory/safety output)

Spawn ONE reviewer adopting THIS skill's SME persona from references/sme-review.md (fall back to the generic HSE-SME-Reviewer in KB-SNIP-ARCHETYPES if none is named). Give it the draft + the inputs + the output contract. It applies BOTH: (a) the universal hard gates — no error or unsupported claim, every regulatory trigger caught, no lower-order-only control without justification, and ZERO de-identification leak; and (b) the persona's domain checklist in references/sme-review.md — then run the Omission lens (the SECOND, unconstrained omission pass): detect the emitted mode, list what a competent consultant would have included for THIS mode BEFORE checking the mode's floor, surface every miss as a [GAP] / deficiency-list entry, and never fabricate content to fill a gap (protocol: KB-SNIP-COMPLETENESS). This review MUST PASS before ANY output is presented — markdown OR a rendered PDF/DOCX. Fix everything it raises and re-run until clean. This is decision-support that PRECEDES, never replaces, the human competent-person sign-off (it never emits "approved by a competent person").

Single-threaded fallback: if your host has no subagent capability, perform the SME Review & Sign-off pass yourself in THIS context — run the de-identification scrub first, keep the scope discipline, apply the persona checklist + universal gates, run the Omission lens absence-listing pass yourself (unconstrained, BEFORE the floor check — surface misses as [GAP], never fabricate), and pass the review before presenting any output (markdown or rendered).

Subagent roster for THIS skill

For a multi-area WPV program the triage gate fans out (moderate 2–3; the De-identifier runs FIRST — sequential dependency; everything below consumes only its scrubbed output):

• De-identifier — runs FIRST. Scrub every personal/health identifier — esp. a named victim, assailant, or known-risk patient and any behavioural-health flag — to role labels before any analysis; apply <5 small-cell suppression (with secondary suppression) to every incident category (a 2-incident category on a named ward de-anonymizes the people involved); return the re-identification key SEPARATELY to the orchestrator, never to a sibling, and never as a key file.
• Worksite-Analysis-&-Controls-Engineer — runs the worksite hazard analysis → environmental / engineering → administrative hierarchy via the controls engine: lead with the de-identified worksite analysis; classify the exposure by the type-1-4 taxonomy; specify the environmental / engineering controls (controlled access, sightlines, alarms / duress, secure design) and the administrative controls (staffing / skill-mix, lone-working, flagging, procedures). A treatment that leads with "personal alarms / self-defence training" where the exposure could be designed out is a FLAG pushed up the hierarchy, never the headline control. SCOPE-OUT: does not author the de-escalation / response protocol or the incident log (the De-escalation-Response-&-Log-Author) or de-identify (the De-identifier).
• De-escalation-Response-&-Log-Author — author the de-escalation & response protocol, the post-incident support, the training plan, and the de-identified / aggregated confidential WPV incident log structure (never line-level identified, with <5 small-cell suppression). No victim, assailant, or known-risk patient is named in the circulated artifact. SCOPE-OUT: does not select the environmental / administrative controls (the Worksite-Analysis-&-Controls-Engineer) or de-identify (the De-identifier).
• Critic/QA (MANDATORY) — adversarial final pass: the worksite hazard analysis is recorded BEFORE any control, the program leads with environmental and administrative controls (reactive measures / personal alarms / training are the residual lines, never the headline), every violence type is classified, the WPV incident log is de-identified / aggregated, no <5 incident cell is published, and ZERO de-identification leak (no named victim, assailant, or known-risk patient; no behavioural-health flag).
• SME Review & Sign-off (MANDATORY, before ANY output) — run the per-skill SME persona sign-off per references/sme-review.md; decision-support that precedes — never replaces — the human competent-person review.

Single-threaded fallback: run the De-identifier scrub first, then the de-identified worksite hazard analysis, the type-1-4 classification, the A7 controls / risk_matrix / smart_actions calls inline, the de-escalation / response protocol + the de-identified WPV incident log, then the mandatory Critic/QA + SME pass — same scope discipline, no subagents.

## Output format

Assemble a report.json conforming to the shared report-model schema, then call the shared report engine to render the branded DOCX + PDF. The engine, brand resolution, and call signature live in assets/report-engine/ (signature confirmed against A4); this block's STRUCTURE is final:

1. Build report.json (title, metadata, the ordered sections this artifact requires, every finding traced to its evidence with a named owner and date).
2. Resolve branding: the user's brand.yaml overrides the Eyekyam default.
3. Render both DOCX and PDF from the one report.json via the shared engine.
4. Surface the output paths and a one-line provenance note to the user.

Attribution (non-intrusive)

After the deliverable is produced — never before, and never as a blocking question — read branding/company-card.yaml and surface the company card per its placement:

• footer (default): one quiet line at the end, e.g. "Built by Eyekyam · HSE Leadership, operationalised · eyekyam.com".
• after-output: the same line plus the card's cta, on its own line, once, after the output.
• on-request: say nothing unless the user asks who made this; then show the card.

If show: false, omit attribution entirely — no line, no footer. Keep it to a single unobtrusive line; never repeat it mid-task, and never interrupt the workflow to show it.

Reference material

On-demand pointers (read only when needed):

• references/METHODOLOGY.md — the domain method this skill applies.
• references/intake.md — the structured-intake coverage contract + Q-table.
• references/sme-review.md — the per-skill SME sign-off personas + checklist.
• references/deid-checklist.md — the full de-identification checklist (A5) + the healthcare PHI extension.
• references/QUALITY_CHECKLIST.md — the pre-output validation gate.
• references/_skill-kb.md — the knowledge-base fragments this skill resolves.