safety-culture-assessment

name	safety-culture-assessment
description	Produces a consultant-grade safety-culture assessment for a named organisation, site, or business unit — designing a survey, mapping results to a maturity model (DuPont Bradley / Hudson Hearts-and-Minds / Westrum) or the Schein three-levels diagnostic lens, triangulating at least two data sources, and emitting a defensible advancement roadmap. Use this skill whenever a user asks to assess, diagnose, benchmark, or measure safety culture or safety climate, run a culture-maturity assessment, design a culture survey, or place an organisation on a Bradley/Hudson/Westrum ladder or against Schein's three levels of culture. It refuses to produce a maturity rating from a single survey (requiring a model, a named population, and >=2 triangulated data sources), suppresses any sub-5 respondent cohort to protect confidentiality, drives a systemic advancement roadmap with named owners and dates, and emits a branded report. Decision-support only; a competent person must review the output.
license	Apache-2.0
metadata	{"author":"eyekyam","version":"1.0","category":"culture","tier":2,"audience":["M","C"],"industry":["All"],"jurisdiction":["All"],"status":"stable","plugin":"hse-leadership","hse_reviewed_by":"","hse_reviewed_date":""}

Safety Culture Assessment

A consultant-grade HSE skill that produces a defensible safety-culture assessment for a named organisation, site, or business unit. It designs (or interprets) a culture survey, maps the result to a recognised maturity model — DuPont Bradley Curve, Hudson / Parker "Hearts & Minds" ladder, or Westrum typology — or, where the user wants a diagnosis rather than a rating, applies the Schein three-levels lens (artifacts · espoused beliefs & values · basic underlying assumptions), and emits a systemic advancement roadmap with named owners and dates.

It forces the single lever that separates a defensible artifact from copy-paste paperwork: it refuses to produce a maturity rating from a single survey — a rating requires a named model, a named population, and at least two triangulated data sources (survey · observation/walk data · records). Respondent confidentiality is protected: any cohort/sub-group breakdown with fewer than 5 respondents is suppressed (with secondary suppression), no finding is attributed to a nameable individual, and no verbatim respondent quote is reproduced in a way that identifies the speaker. Improvement actions are ranked through the full hierarchy of controls (a systemic work-design fix, never "tell people to care more") with a named owner and an ISO due date.

When to use this skill

Use this skill when the user needs a safety-culture assessment for a named organisation, site, or unit — for example "assess the safety culture at the Northgate distribution centre", "where are we on the Bradley Curve / Hudson ladder and how do we advance?", "design a safety-culture survey for the maintenance division", or "run a Schein three-levels diagnosis of why our 'safety first' value isn't enacted on the floor". Trigger phrases: safety culture, safety climate, culture maturity, culture survey, Bradley Curve, Hudson / Hearts & Minds ladder, Westrum typology, Schein three levels, espoused-vs-enacted gap, culture advancement roadmap.

This skill assesses the organisation's culture, not an individual. It is distinct from its siblings — bbs-program-designer (designs a behaviour-observation programme), safety-walk-gemba (structures a single leadership walk), and leading-lagging-kpi-framework (designs the balanced KPI set). Intake-Q1 disambiguates which of these the user actually needs. If the request is vague — "rate our culture" off one pulse survey — the Workflow intake below refuses to produce a rating until a model, a named population, and ≥2 data sources are captured; it will not place an organisation on a ladder from a single survey.

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.

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.

This skill always grounds in the culture-model library, the maturity-band data, the leading-indicator catalogue, and the ISO 45001 leadership clause map — then resolves the jurisdiction only for the underlying legal duty (culture assessment is method-led, not form-led).

Read on every run	File
The four culture lenses — Bradley / Hudson / Westrum maturity ladders + the Schein three-levels diagnostic lens (descriptors + diagnostic probes; Schein is selectable and combinable, not a 4th ladder)	../../knowledge-base/prompt-snippets/culture-models.md (KB-SNIP-CULTURE-MODELS)
Maturity-band descriptors for the ladders + the Schein espoused-vs-enacted GAP rubric (band each model; standalone Schein = a diagnosis, not a rating; quote source+year)	../../knowledge-base/data-points/culture-maturity.md (KB-DATA-CULTURE-MATURITY)
The curated leading/lagging indicator catalogue (so the advancement roadmap is tracked to real leading indicators, not vibes)	../../knowledge-base/data-points/leading-indicators.md (KB-DATA-LEADING-INDICATORS)
ISO 45001 leadership clause cross-walk (5.1 leadership & commitment · 5.4 consultation & participation — the clauses this assessment evidences)	../../knowledge-base/prompt-snippets/leadership-clause-map.md (KB-SNIP-LEADERSHIP-CLAUSE-MAP)
ISO 45001:2018 clause 5.1 / 5.4 (the leadership & worker-participation duties a culture assessment evidences)	../../knowledge-base/standards/iso-45001.md (KB-STD-ISO45001)

Jurisdiction	Read for the legal duty
India	../../knowledge-base/regulatory/in-factories-act.md (+ in-state-forms.md for the user's state — mandatory state detection; defers to hse-india, never a national form number)
UK	../../knowledge-base/regulatory/uk-hswa.md (HSWA 1974 — leadership/worker-participation duties)
USA	../../knowledge-base/regulatory/us-osha.md
EU	../../knowledge-base/regulatory/eu-osh.md (OSH Framework Directive 89/391/EEC — worker consultation)
Unknown	Ask before citing any specific law — the culture models + ISO 45001 5.1/5.4 still apply as the method

Always apply KB-SNIP-HOC to every advancement action: a culture gap is closed by a systemic work-design / leadership-system fix, never by "tell workers to care more". For any maturity band, resolve the model in KB-DATA-CULTURE-MATURITY and quote its source+year — never a remembered cut-off. 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.)

Step 0 — Structured intake (run this first, one question at a time)

The full typed, branched intake — the intake-coverage manifest, the question table, the sibling-disambiguation Q1, the model-selection Q2 (Bradley · Hudson · Westrum · Schein three-levels · combined), the mandatory India→state branch, the echo-back, and the refuse-on-vague anchors — lives in references/intake.md. Run it one question at a time, branch on the answers, echo the captured facts back before any analysis, and refuse to proceed on a vague request. The hard refuse anchors:

1. Q1 (unit + sibling disambiguation) — a named organisation / site / business unit, and a confirmation that the user wants a culture assessment (not a BBS programme → bbs-program-designer, a single leadership walk → safety-walk-gemba, or a KPI set → leading-lagging-kpi-framework). Refuse "the whole company" without a named scope or sampling plan.
2. Q3 (data sources — the triangulation gate) — at least two data sources (validated culture survey · leadership-walk/observation data · records such as reporting/participation/turnover). Never produce a maturity rating on a single survey or a single source. A standalone Schein output is allowed as a diagnosis (named gaps) but still requires the named population and ≥2 sources to be defensible.

Q2 selects the model/lens (and whether Schein is combined with a ladder as a triangulation lens), Q4 captures the confidentiality threshold (default ≥5 respondents per published cohort cell), and Q5 the jurisdiction (India → mandatory state branch).

The culture-assessment method (maturity models + the Schein lens)

Full method in references/METHODOLOGY.md. Steps:

1. De-identify the inputs FIRST — survey responses, focus-group/walk notes, and any records are sensitive personnel data (the deid block above + the reinforced references/deid-checklist.md + the De-identifier-runs-first orchestration rule). Suppress any cohort/sub-group breakdown with <5 respondents, apply secondary suppression, scrub every nameable individual to a role/group label, and never reproduce a verbatim respondent quote that identifies the speaker in a small group. Everything downstream consumes only the scrubbed, aggregated text.
2. Confirm model + named population + ≥2 sources (the refuse gate) — verify a named model/lens (Q2), a named population (Q1), and ≥2 triangulated data sources (Q3) are present. If only a single survey is available, refuse to produce a maturity rating — record [GAP], ask for corroborating observation/records, and offer a Schein diagnosis of named gaps instead of a rating. Do not invent a band.
3. Map to the model (rating) or run the Schein lens (diagnosis) — for a ladder model (Bradley/Hudson/Westrum), band the triangulated evidence against KB-DATA-CULTURE-MATURITY (quote source+year) using the descriptors/probes in KB-SNIP-CULTURE-MODELS. For Schein (selectable as the primary lens, or combinable with a ladder as a triangulation lens), apply the espoused-vs-enacted GAP rubric: compare the espoused values (level 2 — slogans, policy, "safety first") against the underlying assumption enacted in artifacts and behaviour (levels 1 & 3 — what actually wins when schedule and safety collide). A wide gap is a named diagnostic finding, not a maturity band — standalone Schein produces a three-levels diagnosis, never a ladder rating.
4. Locate the systemic drivers — for each band/gap, name what in the leadership system / work design holds the culture where it is (ownership of safety, how information flows — Westrum, how bad news is treated, consultation under ISO 45001 5.4). Attribute to roles/groups and systems, never to a nameable individual. Flag [GAP] where the evidence is thin.
5. Advancement controls (the hierarchy lever) — propose the moves that advance the culture and apply KB-SNIP-HOC, then call controls.rank_controls + controls.validate_treatment. (visible felt-leadership cadence, fixing the consequence system so the safe choice is the easy choice, genuine worker consultation, closing the espoused-vs-enacted gap) rank "awareness campaigns" or "tell people to care more". A roadmap whose only move is a poster campaign is a .

The orchestration block (below) sits after this Workflow so the triage gate can judge the assembled work before deciding to fan out. The control-ranking / action-register steps (5, 6 via controls/smart_actions) are A7 script calls — never a fan-out job.

Sibling cross-reference (emit in the output): "This is a culture diagnosis; to operationalise it use bbs-program-designer (observation programme), safety-walk-gemba (leadership walks), and leading-lagging-kpi-framework (the balanced KPI set that tracks advancement)." — companion skills, D-03 / CONV-12.

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

This is the moderate roster (A6 "moderate = 2–3 fan-out") with the De-identifier as the sequential first gate — critical here because culture-survey responses, focus-group/walk notes and verbatim quotes are sensitive personnel data that can re-identify a respondent in a small cohort. Control-ranking and the SMART advancement register are deterministic A7 script calls (controls, smart_actions) at Workflow steps 5/6 — never an LLM fan-out job. Archetypes: KB-SNIP-ARCHETYPES.

• De-identifier — runs FIRST (sequential gate, not a fan-out peer), and is the most critical job in this skill. Before any analysis: scrub every nameable individual to a role/group label; suppress any cohort/sub-group breakdown with fewer than 5 respondents and apply secondary suppression so a suppressed cell cannot be back-calculated from totals; never reproduce a verbatim respondent quote that identifies the speaker in a small group; ensure no finding is attributed to a nameable individual. Apply references/deid-checklist.md (the reinforced <5 / small-cohort checklist) in full. Return the re-identification key SEPARATELY to the orchestrator — never to a sibling, never in the document. Every job below consumes only this scrubbed, aggregated output. SCOPE-OUT: does not map maturity (Culture-Analyst) or draft the roadmap (Roadmap-Planner).
• Culture-Analyst — map the triangulated evidence to the chosen model via KB-SNIP-CULTURE-MODELS, banding against KB-DATA-CULTURE-MATURITY (quote source+year) across ≥2 sources (never a single survey); for Schein, run the espoused-vs-enacted GAP diagnosis (named gaps, not a band). Names roles/groups and systems, never individuals. SCOPE-OUT: control ranking (the A7 controls call), the SMART roadmap (Roadmap-Planner).
• Roadmap-Planner — turn each accepted advancement move into a SMART action with a named role-label owner + an ISO due date + a measure + a tied leading indicator from KB-DATA-LEADING-INDICATORS (smart_actions.validate_register). SCOPE-OUT: mapping maturity (Culture-Analyst), de-identification (De-identifier).
• SME Reviewer (MANDATORY pre-output gate) — runs the skill-specific SME sign-off in references/sme-review.md (Safety Culture / OD Consultant) before any output: no rating from a single survey (model + named population + ≥2 sources), Schein used as a gap diagnosis not a ladder band, systemic advancement controls above awareness campaigns, no individual named, no <5-cohort cell or identifying quote published.
• Critic/QA (MANDATORY) — adversarial final pass: a named model/lens; ≥2 triangulated sources for any rating; advancement controls ranked through KB-SNIP-HOC (no poster-campaign-only roadmap); every action owned + dated + tied to a leading indicator; culture models cited as method accurately; and ZERO de-identification leak (no residual identifier, no <5 cohort cell, no identifying verbatim quote, no individual named, no re-id key in the output). PASS/FAIL.

A single small unit with one clean data set may run single-threaded — no subagents — but the De-identifier scrub + <5 suppression, the A7 control/action calls, and the Critic/QA + SME passes are still made.

## 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).
• references/QUALITY_CHECKLIST.md — the pre-output validation gate.
• references/_skill-kb.md — the knowledge-base fragments this skill resolves.