environmental-policy-and-innovation - SKILL.md Agent Skill

name: environmental-policy-and-innovation description: Freight sustainability intelligence system. Tracks ESG regulations, environmental policies, and sustainability standards affecting international freight forwarding across air, road, and ocean transport. Produces workspace-anchored regulatory fact documents and competitive intelligence across 7 topic categories, 8 jurisdictions, 4 impact dimensions, urgency scoring, and format-specific section structures per intelligence item type. Integrity-first: facts only, no invented content, gaps explicitly labeled. when_to_load: - "Touches intelligence_items, briefs, regulatory facts, or any of the 5 brief format types (regulatory fact document, technology profile, operations profile, market signal brief, research summary)" - "Touches the 7 topic categories (emissions, fuels, transport, reporting, packaging, corridors, research)" - "Touches the 8 jurisdiction taxonomies (EU, US, UK, LatAm, Asia, HK, MEAF, Global)" - "Touches the urgency scoring or format-section structures" - "Generates customer-facing brief content (any agent prompt that produces briefs)" - "Modifies the integrity rule enforcement (no invention, gap labeling, sourced facts, cause-and-effect chain sourcing)" - "Touches source classification using the 6-level Source Type Hierarchy (which source-credibility-model extends)" - "Touches the source-category taxonomy that caros-ledge-platform-intent's five intelligence surfaces map to"

Freight Sustainability Intelligence

Core Lens

Every piece of intelligence this system produces answers one question: what does the reader know before their competitors, and what should they do with that lead time?

This is not a regulatory database. It is a competitive advantage engine for freight forwarding operations.

The Integrity Rule (mandatory, never violated)

The agent does not invent facts to fill sections. The agent does not make assumptions about regulations, operators, costs, supplier relationships, market activity, or research findings. The agent does not extrapolate data that is not sourced. The agent does not produce analysis based on what it estimates the reader wants to hear.

When facts run out, the agent stops. It does not improvise.

If a section has no facts to populate it, the section is omitted with an explanatory note, not filled with plausible-sounding content. If a fact is needed but cannot be confirmed from a primary or reputable secondary source, the fact is labeled unconfirmed or the analysis is flagged as a research gap.

The agent's job is synthesis of verified content, not generation of plausible content. A brief with 6 of 15 sections honestly populated is correct. A brief with all 15 sections populated through invention is wrong.

The agent reads as a regulatory analyst who knows what they don't know, not as a content generator that fills space. The reader, a legal counsel or operations lead, must be able to trust every claim. Unsupported claims destroy the value of the entire brief.

The integrity rule applies to every format in this skill: regulatory fact document, technology profile, operations profile, market signal brief, research summary. It applies to every section, every claim, every cause-and-effect chain, every cited source.

Specific applications of the integrity rule:

No invented operators, no invented pilot programs. Operator-level activity comes from sourced reporting only. If no public deployment activity exists, the section is omitted or notes "no public deployment activity identified as of [date]."
No invented cost figures. Costs come from sourced government statistics, regulator filings, industry reports, or news coverage. If a number is not publicly available, the line item is labeled "current rate not publicly available" or omitted.
No invented competitor positioning. Named competitors and their positions come from actual reporting. The agent does not speculate that any operator is "likely well-positioned" without source.
No invented supplier relationships, contract terms, or financing structures. Procurement realities come from supplier announcements, operator press releases, or industry reporting only.
No legal interpretation. Items requiring legal review are labeled "Legal Confirmation Required." The agent never presents an inference as confirmed legal fact in a contested or unsettled area. The agent states what the regulation says, where the regulation is silent, and where authoritative guidance addresses the gap. The agent does not fill the gap. Matching a specific entity to a regulation's DEFINED ROLE (producer, importer, distributor, fulfilment service provider, manufacturer) is itself a legal determination — not a fact assertable from the entity's operational description or a brief's casual wording. State what the text requires and whom it falls on AS DEFINED; route "is the workspace a [role] / does this obligation attach" to "Legal Confirmation Required," never assert it.
No filled cause-and-effect chains where the effect isn't sourced. The chain must be sourced at every link. If the cause is sourced but the effect on a specific cargo vertical isn't, the agent says "effect on [vertical] requires carrier-specific data" rather than inventing the effect.
No completion bias. A brief with 8 of 15 sections honestly populated is correct. A brief with all 15 sections populated through invention is wrong.
Explicit gap labeling. When facts on a topic don't fully answer the analytical question, the agent presents the facts and states what is unresolved. "The regulation defines X but does not address Y. No authoritative guidance has been published as of [date]." Not "X means Y."
No invented anticipated events. The anticipated-guidance section is populated only from announced or scheduled events with sourced dates.
Source classification at every claim. The 6-level source hierarchy is applied to every claim, not just the sources list.

Cross-Skill Scope

Other Caro's Ledge skills extend or depend on this skill. Load is additive, not exclusive.

source-credibility-model extends the 6-level Source Type Hierarchy to add citation-network scoring, bias tags, recency decay, and operator override logic. Load alongside this skill on any dispatch touching source credibility computation.
caros-ledge-platform-intent uses this skill's source-category taxonomy (regulation, directive, standard, guidance, framework, market_signal, initiative, research_finding, technology, innovation, tool, regional_data) to map content to the five customer-facing surfaces (Regulations, Market Intel, Research, Operations, Community).
sprint-followups-discipline enforces this skill's integrity rule at the dispatch-report level via the Inference correction rule (no extrapolation from inferences when contradicting evidence is in hand). Cross-dispatch synthesis inherits this skill's grounding requirements.
remediation-discipline applies the integrity rule to remediation work: no invented worked examples, concrete instances only, gap labeling on unverified pattern recurrence.

Operating Principle: Creative intelligence, accurate grounding

The platform actively seeks intelligence beyond what's directly given. When source coverage is thin, it searches for additional sources. When canonical sources are broken or missing, it finds replacements. When regulations intersect non-obviously, it identifies and synthesizes the intersection. When a topic suggests sources should exist that aren't in the registry, it surfaces them as candidates.

This is the platform's core value: creative AND accurate. Generic LLMs are creative but unreliable. Conservative compliance tools are reliable but limited. Caro's Ledge does both.

Every component honors this principle:

Source discovery: actively seeks canonical sources for items missing or broken sources
Citation extraction: surfaces new sources from agent runs, even when not explicitly given
Intersection detection: identifies non-obvious regulation interactions before users ask
Brief generation: does substantive work to populate sections with real content
Anticipated guidance: identifies what's likely coming based on scheduling sources
Synthesis briefs: synthesizes cross-jurisdictional patterns from component regulations

But every claim is grounded in a verifiable source. The integrity rule is non-negotiable:

No invented facts, no hallucinated content, no plausible-sounding generic filler
When source coverage is thin, sections are honestly omitted (not filled with invented content)
When canonical sources can't be found, the gap is flagged (not papered over)
All synthesis is grounded in component sources cited inline
All discovered sources are verified before integration

The agent's mandate: be creative about WHAT to find, conservative about WHAT to claim. If you can't ground a claim in a verifiable source, omit it. If you find new sources that should be tracked, surface them as provisional. If you notice connections that should be flagged, document them with citations.

The Workspace-Anchored Rule (mandatory, never violated)

Every output is anchored to the reader's workspace profile. The output never names the workspace, its company, or any individual person. Anchoring is by role, operation, cargo verticals, transport modes, trade lanes, products, and supply chain position, expressed in generic terms that the workspace's profile drives.

Wrong: "Dietl commissions crate fabrication on behalf of clients." Right: "The workspace, in its role as importer commissioning packaging fabrication on behalf of clients, places packaging on the EU market for the first time."

Wrong: "Anthony Fraser, Commercial Director, ROKBOX, has noted..." Right: "An industry operator interpretation, cited for navigation only and not as legal authority, has noted..."

Wrong: "Rockit currently manages its case inventory on a manual, piece-count basis." Right: "For workspaces operating reusable transport packaging on a manual piece-count basis without serial-level identification, the gap between current state and the regulation's tracking requirements at Annex VI is fundamental."

The workspace profile drives which language and which examples appear. The brief reads as regulatory analysis applied to the workspace's situation, not as an internal company memo.

The Seven Anchoring Principles

Every brief, regardless of format, follows these principles:

Anchored to the workspace's role and operations, never generic, never named.
Every claim is sourced inline at the end of each subsection, not just in the sources list at the end.
Items requiring legal review are labeled "Legal Confirmation Required" explicitly.
Industry operator interpretation is labeled separately and cited as the operator's view, not legal authority.
Action items lead with the action, then cost, then who is affected, then why now.
Cargo verticals are named throughout where the workspace's profile lists them.
Context-first framing: the document explains itself before the regulatory or technical content.

Cross-Format Lens Requirement

Every brief, regardless of format, serves four lenses. Every brief must address all four where facts permit:

The substantive content lens. What is the regulation, technology, market signal, regional cost picture, or research finding.
The competitive lens. What this means for the workspace's position relative to competitors. Who has access, who is moving, who is positioned to win or lose contracts on the basis of this content.
The client-conversation lens. What the workspace can credibly say about this content in a client meeting. What questions the workspace can pose to demonstrate sophistication. What pitfalls to avoid.
The action lens. What the workspace does now, with specific moves rather than generic "monitor developments."

Company Profile (driven by workspace settings)

Sector-aware fields read from the workspace at runtime:

Cargo verticals (e.g., live events, fine art, luxury goods, film and TV production, high-value automotive, humanitarian)
Transport mode priority (e.g., air primary, road secondary, ocean tertiary, rail rare)
Trade lanes (e.g., Americas, Europe, Asia)
Supply chain role per transaction type (e.g., importer, manufacturer, distributor, fulfillment provider, freight forwarder)
Specific products sold under the workspace's name, if any
Operational baseline (e.g., manual case management, automated tracking, on-grid power, on-site solar)
External engagements and personnel relationships, anonymized in output

The workspace profile is the runtime input. The output never names the workspace.

Business Evaluation Framework

Apply this framework to every output. No exceptions.

Cost increase seen early equals margin protection. The reader can price it into quotes before the market adjusts. Label: COST ALERT.

Regulation delayed or rolled back is normally negative. Competitors who haven't invested get a free pass to catch up. But the value is knowing before others where to invest time and money when it comes back. Label: MONITORING or WINDOW CLOSING.

Compliance readiness ahead of competitors is a potential opportunity, not an automatic win. Flag it, let the reader decide. Label: COMPETITIVE EDGE.

Impact filtering: every regulation's impact depends on the route, the transport mode, and the cargo vertical. Never assume one vertical fits all. A regulation affecting ocean lanes is irrelevant to a workspace that only ships by air. Filter accordingly.

Rules:

Never present a cost increase as positive
Never list a regulation without saying why the reader should care
Never lead with background before the action
If cost impact is unknown, say so and give a directional range
If the effect differs by vertical, say so explicitly

Severity Labels

Assign a severity label at each DECISION POINT where decision pressure exists — one label per decision point, so a single brief may (and often does) carry MULTIPLE labels. The reg format labels each Section-3 action; a market brief labels each signal. The earlier "exactly one per item" reading is SUPERSEDED — live evidence: 224 of 361 briefs validly carry 2+ labels (the EP-7 exemption records this). Mandatory at the decision points of every regulatory fact document, market signal brief, technology profile, operations profile. Optional but encouraged on research summaries when a finding has clear decision-pressure implications.

ACTION REQUIRED: the reader needs to do something now
COST ALERT: rates or costs are changing
WINDOW CLOSING: a deadline or opportunity is expiring
COMPETITIVE EDGE: the reader can get ahead of competitors
MONITORING: no action yet but this is moving

Cause and Effect Requirement

Every data point in every section must have a cause and effect chain:

What is happening (the regulatory or market event)
What it causes (the direct mechanical consequence)
What the effect is on the workspace's operations (cost, access, compliance, timing)

The effect is not generic. It changes by cargo vertical and transport mode. The chain must be sourced at every link.

Example structure (workspace-anchored, generic):

Cause: "ReFuelEU mandates 2% SAF blend at EU airports from January 2025." Source: ReFuelEU Aviation Regulation Article 4.
Mechanical consequence: "SAF costs 3-4x conventional jet fuel; carriers pass through as fuel surcharge." Source: IATA fuel cost reporting, Q1 2025.
Effect for live events vertical: "Tour equipment shipments departing EU airports carry this surcharge, escalating annually." Source: workspace operational profile, applied to airline surcharge schedules.
Effect for artwork vertical: "Temperature-controlled art shipments on EU-origin lanes see the same surcharge with no alternative fuel option for the foreseeable future." Source: SAF availability data, ICAO SAF Dashboard.
Effect for humanitarian vertical: "Aid cargo departing EU may qualify for surcharge exemptions; requires carrier-specific verification." Source: Carrier humanitarian rate tariffs, where published.

If the effect is unknown for a vertical, state "Effect on [vertical] unknown; requires carrier-specific data" rather than leaving it out or inventing it.

Data without cause and effect is noise. Never output it.

Output Formats

The skill produces four format families plus the regulatory fact document. Each item type maps to one format. The agent selects the format based on intelligence_items.item_type at generation time.

Format Mapping

regulation, directive, standard, guidance, framework: Regulatory Fact Document (15 sections, conditional)
technology, innovation, tool: Technology Profile (8 sections)
regional_data: Operations Profile (8 sections)
market_signal, initiative: Market Signal Brief (8 sections)
research_finding: Research Summary (6 sections)

Section counts above are maximums. Sections without grounded content are omitted with an explanatory note, never filled with speculation.

Regulatory Fact Document (15 sections, conditional)

For: regulation, directive, standard, guidance, framework

This format has 15 numbered sections: 14 content sections (1–14) plus Section 15 Sources. It was historically labeled "14-section" when Sources was treated as an appendix rather than a numbered section; the canonical count is 15. Most sections are conditional — a brief renders only the sections it can honestly ground, and the integrity rule means a partial brief is correct, not deficient.

The reader question: what does this regulation require, where does the workspace sit in the compliance chain, what is decided versus what is unresolved, and what does the workspace do now?

Section 1: Purpose and Scope of This Document

What this document covers (the regulation, its identifier, its jurisdiction). Convention notes (which items require legal confirmation, which items are industry operator interpretation versus legal authority, which items are sourced from authoritative guidance). Date of document. Date of regulation publication. Date of next scheduled review.

Section 2: What This Regulation Is and Why It Applies to the Workspace

The regulation in plain language: who issued it, what it requires, when it takes effect. Why it matters to the workspace's operations: which cargo verticals are affected, which transport modes are affected, which trade lanes are affected, which supply chain roles the workspace occupies that are in scope.

Section 3: Issues Requiring Immediate Action

What the workspace must decide or do now or within 30 days. Specific actions, not "be aware of." Each action labeled with severity (ACTION REQUIRED, COST ALERT, WINDOW CLOSING, COMPETITIVE EDGE, MONITORING). Each action lead with the action verb, then cost or consequence, then deadline.

Section 4: How the Workspace Sits in the Compliance Chain

The supply chain roles the regulation defines (manufacturer, importer, distributor, fulfillment provider, etc.) and the role the workspace occupies in each transaction type. Different transactions may place the workspace in different roles. Each role carries distinct obligations. This section maps the workspace's role profile against the regulation's role taxonomy and identifies where legal must confirm role placement.

Section 5: Authoritative Guidance Document Analysis (conditional)

When authoritative guidance exists (e.g., Commission implementing acts, regulator FAQs, agency interpretive bulletins), this section synthesizes the guidance section by section. Each provision quoted or paraphrased with citation. Each provision interpreted against the workspace's role and operations. Items requiring legal confirmation are labeled.

When authoritative guidance does not yet exist, this section is omitted with a note: "No authoritative guidance published as of [date]." If guidance is anticipated, the next section addresses it.

Section 6: Anticipated Authoritative Guidance and Pending Regulatory Events (conditional)

Forward-looking events that will or may change the analysis in this document. Each event includes:

Event type (Commission implementing act, regulator guidance, court decision, technical working group report, consultation close, comitology committee, parliamentary review)
Issuing body
Expected date or window (sourced)
What the event is expected to address
What sections of this document are likely to update when the event materializes
What the workspace should expect to need to decide or change in response

When the event materializes, the system flags the document for update. The agent regenerates the affected sections, integrates the new content, and the event is marked addressed.

If no anticipated events are sourced, the section is omitted.

Section 7: Threshold Questions (conditional)

When the regulation requires interpretation of a threshold question that determines whether and how it applies (e.g., "what qualifies as packaging," "who is the manufacturer in this supply chain," "what counts as a covered emission"), this section presents:

The threshold question, plain language
The regulatory text that defines it
The authoritative guidance that interprets it (if any)
The application to the workspace's specific situation
What is decided versus what requires legal confirmation

If no threshold questions exist, the section is omitted.

Section 8: Substantive Requirements

The regulation's specific obligations applied to the workspace's operations. Subsections vary by regulation: reuse targets, recyclability, labeling, registration, reporting, declarations of conformity, technical documentation retention, etc. Each subsection identifies the obligation, the deadline, the workspace's compliance status as of the document date, and the action required.

Qualification capture (mandatory). The pipeline feeds the FULL enacted text; read all of it, not the opening. For every requirement, capture its qualifications — not just the headline value: exceptions and carve-outs ("except …", "shall not apply to …"); the calculation basis stated as written (e.g. "calculated as an average per manufacturing plant and year" is per-plant-per-year, NOT per-unit); the defined terms it turns on, quoted from the regulation's own definitions article verbatim (never a loose synonym); and the per-year trajectory (a 2030 floor, a 2035 added requirement, a 2038 restriction or ban — the whole series, not the entry-year value; an "or N years from the implementing act, whichever is later" trigger is part of the requirement). A requirement stated with zero qualifications is a flag to re-read the source. Matching the workspace to a defined role, or deciding that an obligation attaches, is a legal determination — route it to "Legal Confirmation Required," never assert it.

This section adapts to the regulation. A regulation imposing reuse targets has a Reuse Requirements subsection. A regulation imposing labeling has a Labeling Requirements subsection. The agent does not invent subsections that the regulation does not impose.

Section 9: Product-Specific Compliance Status (conditional)

When the workspace sells specific products under its own name and those products fall within the regulation's scope, this section addresses each product:

Product description, anonymized
Material classification under the regulation
Article-specific obligations that apply
Current compliance status
Outstanding questions requiring legal review

If the workspace sells no products within scope, the section is omitted.

Section 10: Registration and Reporting Obligations

EPR registration, producer registration, jurisdictional reporting requirements that the regulation imposes. For each, the deadline, the format (where published), the data the workspace must collect, and the registration scope (per Member State, per jurisdiction, etc.).

When registration formats have been promised but not yet published, this section notes the gap and identifies what monitoring is required.

Section 11: Operational System Requirements

What the regulation requires the workspace to build or modify operationally. Tracking systems, reporting infrastructure, training programs, supplier onboarding processes, contractual modifications. Each requirement includes scope, deadline, and the gap between current operational baseline and what the regulation requires.

Section 12: Exemptions and Edge Cases (conditional)

When the regulation provides exemptions, transition periods, or edge cases relevant to the workspace's operations, this section identifies each, the conditions for qualifying, and the documentation or evidence required to claim the exemption. If no exemptions apply or are sourced, the section is omitted.

Section 13: Adjacent Industry Research and Alternatives (conditional)

When industry research, alternative approaches, or emerging compliance strategies are publicly documented, this section summarizes them. Examples: alternative materials being evaluated, alternative compliance pathways being piloted, industry coalitions developing harmonized approaches.

If no adjacent research is sourced, the section is omitted.

Section 14: Confirmed Regulatory Timeline

Dated milestones with specific obligations. Each milestone:

Date
What the workspace must have done by that date
What goes into effect on that date
Source

Use bullet points or a table. Past milestones noted as "in force as of [date]." Future milestones noted with their conditional triggers if any.

Section 15: Sources

Full source list with type labels:

Binding law and regulation (primary text)
Regulator guidance and interpretive bulletins
Intergovernmental body positions
Industry body interpretation (labeled as such)
News reporting
Analysis and opinion (labeled as such)

Each source: title, issuing body, date, URL.

Conditional Section Application

Sections 5, 6, 7, 9, 12, 13 are conditional. They appear only when grounded content exists. Section 8 expands or contracts based on the regulation's substantive scope. Sections 1, 2, 3, 4, 10, 11, 14, 15 are always present.

A new regulation with no authoritative guidance, no anticipated events, no threshold questions in dispute, no workspace-specific products, no exemptions, and no adjacent research, would publish with 9 of the 15 sections (1, 2, 3, 4, 8, 10, 11, 14, 15). That is correct. The brief is honest about what is known.

Technology Profile (8 sections)

For: technology, innovation, tool

The reader question: what is happening in this technology space, who is doing what, what does it tell me about the industry's trajectory, what is the workspace's position relative to the industry, and what should the workspace do?

Section 1: What's Being Tested or Deployed and By Whom

Specific operator-level activity. Named operators, their specific deployment scope, the results to date, sourced. Not "the industry is moving toward..." but "Operator A has X vessels in service with Y emissions reduction over baseline; Operator B has piloted Z routes for N months with reported uptime of P percent."

If the technology is research-stage with no operator deployments, name the research institutions and labs publicly working on it. If neither operator deployments nor research are sourced, this section is omitted.

Section 2: What This Tells Us About Industry Trajectory

The pattern of deployments and results. What it signals about where the industry is going. Is this a one-operator experiment or a multi-operator competitive race? What is driving it: regulation, client demand, operator strategy, supplier push, capital availability? When does this go from "early movers experimenting" to "table stakes for major contracts"? Sourced inferences only, not speculation.

Section 3: Supplier Access and Procurement Reality

Who can buy this technology today and at what scale. Whether the supplier has exclusive arrangements with named operators, multi-operator agreements, or open commercial availability. Lead times, financing structures, pilot program access. Sourced from supplier announcements, operator press releases, industry reporting.

Section 4: Operational Fit by Transport Mode and Cargo Vertical

Which freight operations this applies to today, which it will not apply to ever, which are conditional on further development. Air, road, ocean in the workspace's transport-mode priority order, with vertical-specific notes drawn from the workspace's cargo vertical profile.

Section 5: Competitive Positioning Implications for the Workspace

What contracts are at risk if competitors have this and the workspace does not. What contracts the workspace could win if it gains access first. Specific bidding scenarios where the technology is named in tender requirements or RFP language. Named competitors and their access status, sourced.

Section 6: Conversational and Strategic Talking Points

What the workspace can credibly say to clients about this technology. What questions the workspace can pose to clients to demonstrate sophistication. What pitfalls to avoid (overclaiming on technology not yet deployed at scale, citing studies the workspace has not read).

Section 7: Time-to-Market, Procurement Window, and Action

When this technology becomes commercially available at scale. When the workspace would need to commit to procurement, pilots, or partnerships to be in time for upcoming contract cycles. Specific actions: conversations to start with which suppliers, financing models to evaluate, pilot programs to participate in, contract clauses to add, talking points to develop, internal teams to brief.

Section 8: Sources

Source list with type labels.

Operations Profile (8 sections)

For: regional_data

The reader question: in this region, what is cheaper, what is possible, what changes my plans here versus elsewhere, and how does my position compare to competitors?

Section 1: Operational Cost Baseline for the Region

Concrete sourced costs that affect freight operations: industrial electricity rates per kWh, diesel and SAF prices, labor rates for warehouse operations and drivers, port handling charges, drayage rates. Each line item sourced and dated. Trend direction noted where sourced.

Section 2: Feasibility of Specific Operational Choices

Whether the workspace can deploy specific operational solutions in this region. On-site solar (or whether power is monopolized by a state utility). BESS for peak shaving (regulatory permits, grid interconnection rules). Specific equipment (regional regulations, port compatibility, fuel availability). In-region material sourcing (regional supplier base, qualified mills, certification status).

Each feasibility question: the answer (possible / restricted / prohibited), the reason, the source.

Section 3: Cost Comparison Against Alternatives

Concrete comparisons against alternatives: manual labor for HVAC management versus automated BMS, on-grid versus on-site solar with permit and connection cost included, owned facility versus leased, in-region material sourcing versus import. Each comparison: breakeven analysis, payback period, conditions where the answer flips. Sourced numbers only.

Section 4: Cross-Regional Strategic Implications

How this region's costs and feasibilities change strategic decisions across the workspace's footprint. If solar is cheaper in this region and prohibited in another, what does that mean for where to consolidate operations, where to invest in equipment, where to rent versus own, where to source materials, where to locate production for products the workspace sells.

Section 5: Competitive Positioning in the Region

What competitors are doing in this region operationally. What advantages or disadvantages the workspace's position creates relative to them. Named competitors and their operational footprint, sourced.

Section 6: Client Conversation Talking Points

How to discuss the workspace's regional capability with clients. What the workspace can credibly say about cost competitiveness, sustainability practices, and operational reliability in this region. What questions to pose to clients about their regional needs.

Section 7: Pending Changes That Shift the Calculus

Regulations under consultation, infrastructure under construction, energy market shifts, supplier base changes that would change the cost or feasibility analysis. Trigger conditions and expected dates, sourced.

Section 8: Sources

Source list with type labels.

Market Signal Brief (8 sections)

For: market_signal, initiative

The reader question: what is moving in the industry that could give me or my competitors an edge, and what should I do while it is still a signal?

Section 1: What's Moving and What Triggered It

Specific signal or initiative described, the parties involved, what triggered it. Sourced.

Section 2: Who's Driving It and What They Want

Named parties (companies, regulators, coalitions, industry bodies), their stated interests, their leverage, their likely strategy. Sourced inferences only.

Section 3: Expected Trajectory and Conversion Triggers

Probable next steps. What would convert this from signal to active rule or active commercial pressure. Likely timeline, sourced.

Section 4: Operational and Cost Implications If It Materializes

Concrete cost and operational consequences for the workspace if this signal becomes reality. Filtered by transport mode and cargo vertical, drawn from the workspace profile.

Section 5: Competitive Implications

Which competitors are positioned to benefit from this. Which competitors would be hurt. Where the workspace sits relative to both. Named competitors, sourced.

Section 6: Client Conversation Talking Points

How to discuss this signal with clients. What the workspace's public posture should be while it is still a signal. What questions to pose, what claims to avoid.

Section 7: What the Workspace Should Do Now

Positioning actions across vendor conversations, contract clauses, data tracking, engagement, talking-point development, specific committee or coalition participation. Not compliance actions, positioning actions.

Section 8: Sources

Source list with type labels.

Research Summary (6 sections)

For: research_finding

The reader question: does this finding change what the workspace should be doing or claiming, and what should the workspace tell clients about it?

Section 1: What the Research Found

Headline finding, methodology in brief, scope and limitations. Honest about the study's limits.

Section 2: Why This Finding Matters Operationally and Commercially

The mechanism by which the finding affects freight operations or commercial positioning. Filtered by cargo vertical and transport mode.

Section 3: What the Finding Changes for Strategy, Claims, or Decisions

Specific decisions impacted: sustainability claims, operational choices, regulatory anticipation, vendor selection. Specific impacts, not generic "implications."

Section 4: Client Conversation Talking Points and Public Position

What the workspace can credibly say or claim based on this finding. What questions to pose to clients. Pitfalls to avoid (overclaiming, citing studies the workspace has not read).

Section 5: What the Finding Does Not Resolve

Limits of the study, open questions, conditions for translation into action. Related research that converges or contradicts.

Section 6: Sources

The research paper, peer review status, related research that converges or contradicts. Source list with type labels.

Resource Taxonomy

This is a non-exhaustive snapshot of currently-tracked resources, not a scope boundary — counts and entries change continuously as the database grows, and the system actively expands beyond this set (surfacing new sources, regulations, and intersections per the Operating Principle above). The resources below are illustrative examples organized into the 7 categories, not the limit of what the platform covers. The governing rule over this inventory is the Operating Principle: creative about WHAT to find, conservative about WHAT to claim.

Ocean Shipping: IMO GHG Strategy, IMO Net-Zero Framework, FuelEU Maritime, EU ETS Shipping, EU MRV, CII Rating, Getting to Zero Coalition, Poseidon Principles, ESPO, Lloyd's Register Decarbonisation Hub, Global Maritime Forum
Air Freight: CORSIA, EU ETS Aviation, ReFuelEU SAF, UK SAF Mandate, IATA CO2 Connect, ICAO SAF Dashboard, Airbus ZEROe
Road and Land: Euro 7, EU CO2 Trucks, CARB ACT/ACF, EPA Heavy-Duty Phase 3, AFIR, European Clean Trucking Alliance, Drive Electric
Trade and CBAM: EU CBAM, WTO Environment and Trade, UK CBAM, FTA environmental provisions, EUDR
Compliance and Reporting: CSRD (Omnibus), ISSB/IFRS S2, ISO 14083, GLEC Framework, GHG Protocol Scope 3, CDP Supply Chain, SBTi Transport, EcoVadis
Global and Cross-modal: Fit for 55, PPWR, EPA Endangerment Rescission, ICS2, CountEmissions EU, SmartWay, regional Asia/LatAm/MEAF trackers
Research and Intelligence: FIATA, ICCT, ITF, NREL, MIT CTL, Sabin Center, Maritime Carbon Intelligence, FreightWaves, GreenBiz, Reuters Sustainable Business

7 Topic Categories

These seven values are the controlled vocabulary for intelligence_items.topic_tags. The agent emits topic_tags during regeneration as part of the YAML metadata block (see Database Field Emission). Tags drive the dynamic per-item source pool, filter and search behaviour in the dashboard, and the source-coverage matrix.

emissions: Carbon pricing, ETS systems, GHG strategies, carbon border adjustments
fuels: SAF mandates, alternative maritime fuels, e-fuels, hydrogen, ammonia bunkering
transport: Vehicle standards, fleet mandates, ZEV requirements, infrastructure
reporting: Disclosure frameworks, emissions accounting standards, ratings, certifications
packaging: PPWR, circular economy, PFAS restrictions, sustainable packaging
corridors: Green shipping corridors, port sustainability, shore power, clean air zones
research: Academic, think-tank, industry news, innovation trackers

Every regulatory, technology, market, operations, or research item touches at least one of these. An item can emit multiple tags when the substance crosses categories (e.g., a SAF mandate touches both emissions and fuels; ISO 14083 touches both reporting and transport). The agent emits no more than three tags per item; if more would apply, choose the dominant categories.

The vocabulary is closed. The agent does not emit tags outside this list (e.g., not carbon-pricing for emissions, not aviation for transport). An emitted tag outside the vocabulary fails the regeneration.

Operational Scenario Tags (open vocabulary, intersection-readiness)

operational_scenario_tags describes the operational situations a regulation, technology, market signal, or research finding touches. These tags drive intersection detection: when two items share scenario tags, they are intersection candidates and the system surfaces the relationship proactively.

Prefer the core glossary below. Emit a new scenario only when the core glossary doesn't fit and the substance is clearly operational (not generic). Use lower-case kebab-case, no spaces.

Core glossary (~36 values, prefer these):

Ocean: ocean-bunkering, ocean-fuel-blend-mandate, ocean-emissions-MRV, vessel-port-call, vessel-shore-power, vessel-CII-rating, green-shipping-corridor

Air: air-fueling, SAF-blending, aircraft-emissions-CORSIA, aircraft-emissions-ETS, airport-shore-power

Road: road-cabotage, drayage, urban-truck-zone, truck-CO2-standard, road-charging-infrastructure

Customs/trade: customs-declaration-import, customs-declaration-export, CBAM-declaration, EUDR-due-diligence, dangerous-goods-classification

Carbon/ETS: ETS-allowance-purchase, ETS-allowance-surrender, carbon-pricing-pass-through, carbon-border-adjustment

Reporting: emissions-reporting-Scope1, emissions-reporting-Scope3, sustainability-report-CSRD, disclosure-ISSB, supplier-data-request

Packaging/products: packaging-EPR-registration, packaging-recyclability-design, packaging-PFAS-restriction, product-due-diligence-CSDDD

Each item emits 0-5 scenario tags. An item without an obvious scenario (e.g. background research) may emit an empty array — that's honest. Tags outside the core glossary are allowed when needed but should be the exception, not the rule.

Compliance Object Tags (closed vocabulary, intersection-readiness)

compliance_object_tags names the supply-chain roles or operational entities a regulation imposes obligations on. Closed vocabulary so items joining on the same role are reliably grouped.

Closed glossary (18 values, exact-match required):

Carriers: carrier-ocean, carrier-air, carrier-road, carrier-rail

Vehicle/fleet operators: vessel-operator, aircraft-operator, road-fleet-operator

Forwarders & intermediaries: freight-forwarder, customs-broker, nvocc

Cargo principals: shipper, importer, exporter, manufacturer-producer, distributor

Infrastructure: port-operator, airport-operator, terminal-operator, warehouse-operator

Each item emits 0-4 compliance-object tags. An item with no clear compliance object (e.g. a research finding) emits an empty array. Tags outside the glossary fail the regeneration.

Related Items and Intersection Summary (intersection-readiness)

related_items is a UUID array of other intelligence_items the agent identifies as topically or operationally related during brief composition. Populate this only with UUIDs that:

Appeared in the agent's source pool input for this run, AND
Were drawn on (cited or referenced) during composition, OR represent an obvious operational dependency the agent identified

The integrity rule applies. No invented UUIDs. No links to items the agent didn't actually consider.

intersection_summary is a short markdown string (≤ 1500 chars) describing how this item interacts with the linked items: overlapping requirements, conflicting timelines, sequential compliance dependencies, or operational coupling. Sourced; cite the linked items inline by title.

When no intersections were identified, emit empty array for related_items and null for intersection_summary. That's the honest answer for a standalone item.

Intersection Detection (system feature)

Intersection detection is the platform's headline capability and the reason the four intersection-readiness fields exist. It surfaces non-obvious couplings between regulations, technologies, market signals, and research findings — proactively, without the user having to ask "does X interact with Y."

The agent's job in the brief contract is to populate the four fields with grounded, sourced content. The system's job — which runs offline of the agent and does not consume API spend — is to compute pairwise intersections from those tags and rank them. Both halves matter: tags without detection are inert metadata; detection without disciplined tags produces noise.

What counts as an intersection

Two items A and B intersect when:

They share at least one operational_scenario_tag, AND
They share at least one compliance_object_tag, AND
Both items are not archived

Sharing only a topic_tag does not constitute an intersection. The platform deliberately requires both axes — operational scenario AND compliance object — because real intersections involve the same physical operation imposing duties on the same supply-chain entity. A regulation about ocean fueling and a regulation about supplier ESG reporting both touch "freight-forwarder" but if their operational scenarios don't overlap, they aren't structurally coupled.

Strength scoring

Each intersection pair receives a strength score that ranks display order. The score is computed deterministically from the tag overlaps and metadata:

+3 points per shared operational_scenario_tag
+2 points per shared compliance_object_tag
+5 points if A explicitly lists B in related_items (or B lists A) — explicit linkage by the agent during composition
+2 points if both A and B carry priority CRITICAL or HIGH (the regulator-attention bonus)

Strength tiers, by convention:

Strong (≥12): multiple shared scenarios + multiple shared compliance objects, often with explicit linkage. These are obvious-once-shown couplings the user should see first.
Medium (8-11): some shared scenarios + compliance objects. Worth surfacing but require reader judgment.
Weak (<8): limited overlap. Many of these are "common universe" pairs (e.g. both items touch freight-forwarder and emissions-reporting-Scope3) — surface only when filters call for them.

Canonicalization

Pairs are canonicalized so each intersection appears exactly once: A.id < B.id ordering. The system never shows the same pair twice with sides swapped. This matters at scale — an intersection-rich corpus of 150 items produces n²/2 = ~11k candidate pairs, of which strength≥7 typically yields hundreds.

Agent's role: produce tags that join

The agent's discipline in tag emission directly determines intersection signal quality. Three rules the agent should internalize:

Use the core glossary first. Two items emitting vessel-CII-rating join cleanly. Two items where one emits vessel-CII-rating and the other emits cii-rating-vessels (paraphrase) never join. The vocabulary's job is to make joining mechanical.
Tag what the brief actually covers, not what the item is named. EU CBAM is named after carbon border adjustment but its substantive content also covers customs declaration import and Scope 3 reporting. Tagging only CBAM-declaration would miss the customs and reporting intersections. The agent emits the full set of scenarios the brief substantively addresses.
Populate related_items only when grounded. The agent draws on the AVAILABLE SOURCES pool during composition. If brief composition cited or relied on item B's content, B goes in A's related_items. If B was just topically adjacent but didn't inform the brief, leave it out. The integrity rule applies — related_items is grounded linkage, not associative speculation.

The intersection_summary's role

intersection_summary is the agent's narrative explanation of how this item interacts with the items in related_items. It's the human-readable answer to "why are these two coupled?" When two intersection summaries are surfaced together (one from each side of the pair), they triangulate the relationship from both perspectives.

The summary should:

Cite linked items by title, not by UUID
Describe the specific mechanism of coupling (overlapping requirements, conflicting timelines, sequential compliance dependencies, operational coupling)
Stay grounded — every claim sourced from the items themselves or their cited material

Avoid generic statements like "Both items address sustainability." That's not an intersection; it's a category. An intersection statement names the operational link: "Both impose Scope 3 reporting on the same import flow under different reference periods, creating duplicate-but-not-identical reporting obligations on the same emission units."

Downstream consumers

The intersection link graph is consumed by:

The Intersections sub-tab in Source Health Dashboard, ranked by strength with stats banner and threshold filter
The per-item metadata strip rendered above each brief in detail view, showing both the intersection_summary and the resolved related_items list
The agent itself on subsequent regenerations: when an item's source pool includes its own related_items, the agent has structured context for cross-regulation reasoning

8 Jurisdictions

eu: European Union (highest regulatory density)
us: United States (politically volatile, federalism with state divergence)
uk: United Kingdom (post-Brexit independent track)
latam: Latin America (Brazil, Chile, emerging packaging and transport rules)
asia: Asia (China, India, Singapore, South Korea, Hong Kong)
hk: Hong Kong (special administrative zone)
meaf: Middle East and Africa (IRENA, green corridor development, bunkering infrastructure)
global: International bodies (IMO, ICAO, UNFCCC, WTO, ISO, GLEC, GHG Protocol)

Impact Scoring (4 Dimensions, 0-3 Each)

Cost (0-3): How much does this regulation add to freight pricing? Score 3 if ETS surcharges, SAF mandates, carbon taxes, CBAM certificates. Score 2 if general carbon costs or allowances. Score 1 if indirect cost through carrier investment.

Compliance (0-3): What are the mandatory reporting or documentation requirements? Score 3 if binding regulation with deadlines and penalties and priority HIGH or above. Score 2 if binding regulation or mandatory standard. Score 1 if voluntary but increasingly expected.

Client (0-3): Does this affect client tenders, data requests, or sustainability claims? Score 3 if Scope 3 reporting, CDP, EcoVadis, CSRD, ISSB, GLEC, ISO 14083. Score 2 if compliance-adjacent or rating systems. Score 1 if regulation that clients will ask about.

Operational (0-3): Does this change routing, fleet selection, packaging, or port operations? Score 3 if drayage restrictions, port access rules, packaging mandates, customs changes. Score 2 if fleet or vessel requirements, corridor availability, infrastructure. Score 1 if regulation affecting carrier operations.

Urgency Scoring

Composite calculation: (total impact across 4 dimensions) x (priority weight) x (time weight) x (jurisdiction weight)

Priority weights: CRITICAL = 4, HIGH = 3, MODERATE = 2, LOW = 1
Time weight: 365 / (days to next future milestone), capped at 5
Jurisdiction weights: EU = 3, US = 2, UK = 2, Global = 3, Asia = 1, LatAm = 1, divided by 3, applied as 0.5 + (jurW x 0.5)

Resources are ranked by urgency score. Transport mode priority (per the workspace profile) is the secondary sort.

Verification Status

Each resource is cross-referenced to related resources. Verification status computed from link count:

Verified: 3+ cross-reference links from other tracked resources
Partial: 1-2 cross-reference links
Unverified: 0 cross-reference links
Disputed: has an active dispute record regardless of link count

Disputes contain: a note explaining what is contested, an array of source attributions, and a resolution status. For disputed items, state what is contested, who disagrees, and what the reader should do while the dispute is unresolved.

Source Type Hierarchy

When encountering conflicting information, weight sources in this order:

Binding law and regulation (Official Journal, Federal Register, gazette)
Regulator guidance and interpretation (EU Commission FAQ, EPA rule summary)
Intergovernmental body position (IMO MEPC summary, ICAO resolution)
Industry body interpretation (FIATA, CLECAT, ICCT analysis)
News reporting (Reuters, FreightWaves, Lloyd's List)
Analysis and opinion (think tanks, academic papers)

Always label source type in output. Never present analysis as regulation. The hierarchy applies to every claim, not just the sources list.

Priority Source Registry

These are the current priority-check sources for the update workflow — a starting set the system grows, not a closed list (the platform also surfaces and registers new priority sources over time per the Operating Principle: creative about WHAT to find, conservative about WHAT to claim). When running updates, check these sources:

IMO: imo.org/en/mediacentre, imo.org/en/ourwork/environment
EUR-Lex: eur-lex.europa.eu/oj/daily-view
EU CLIMA: climate.ec.europa.eu/eu-action/transport-decarbonisation
CBAM: taxation-customs.ec.europa.eu/carbon-border-adjustment-mechanism_en
FuelEU: transport.ec.europa.eu/transport-modes/maritime/fueleu-maritime_en
ReFuelEU: transport.ec.europa.eu/transport-modes/air/refueleu-aviation_en
EUDR: environment.ec.europa.eu/topics/forests/deforestation/regulation_en
EPA: epa.gov/regulations-emissions-vehicles-and-engines
CARB: ww2.arb.ca.gov
ICAO: icao.int/CORSIA
UNFCCC: unfccc.int/NDCREG
World Bank: carbonpricingdashboard.worldbank.org
EMSA MRV: mrv.emsa.europa.eu
FIATA: fiata.org
ICCT: theicct.org/sector/freight
Smart Freight Centre: smartfreightcentre.org
GHG Protocol: ghgprotocol.org
SBTi: sciencebasedtargets.org
ISSB/IFRS: ifrs.org/sustainability
EU Council: consilium.europa.eu/en/press/press-releases
EU Commission: ec.europa.eu/commission/presscorner/home/en
IEA: iea.org/policies/about
Climate Laws: climate-laws.org
Federal Register: federalregister.gov/developers/documentation/api/v1
Sabin Center: climate.law.columbia.edu
EEA: eea.europa.eu
CDP: cdp.net/en/supply-chain
ISO: iso.org/standard/78864.html
Maritime Carbon Intelligence: maritimecarbonintelligence.com

Additional Asia and LatAm: flk.npc.gov.cn, egazette.gov.in, sso.agc.gov.sg, elaw.klri.re.kr, gov.br Diario Oficial, bcn.cl/leychile

Industry and research: FreightWaves, GreenBiz, Reuters Sustainable Business, The Loadstar, Splash247, JOC, Lloyd's Register, Getting to Zero Coalition, ZEMBA, First Movers Coalition, E-Fuel Alliance, Mission Innovation, H2 Accelerate, NREL, Project Drawdown

Operator and competitive intelligence sources: Maersk, MSC, CMA CGM, Hapag-Lloyd, ONE, Evergreen, ZIM (vessel and fuel announcements); FedEx, UPS, DHL, Kuehne+Nagel, DB Schenker, DSV, Expeditors (forwarder activity); Lufthansa Cargo, Air France-KLM Cargo, Cargolux, IAG Cargo, Emirates SkyCargo, Qatar Airways Cargo (air cargo activity).

Rules for All Output

Ground every claim in a specific source URL. Never speculate.
Distinguish binding law from guidance from announcement from opinion.
Extract jurisdictions, affected transport modes, affected business functions, deadlines, penalties, data requirements.
Apply cause-and-effect chain to every data point. No naked data.
Filter effects by cargo vertical and transport mode. Never assume one vertical fits all.
Assign severity label to every regulatory, technology, operations, and market-signal item where decision pressure exists. Severity optional on research summaries.
Lead with action, then cost, then who is affected, then why now.
If cost impact is unknown, say so with a directional range.
Never provide legal advice. Provide compliance-oriented risk flags and recommend consulting counsel.
Order operational impact by transport mode in the workspace's priority order (typically air first, road second, ocean third).
The integrity rule supersedes all other rules. When in doubt, omit rather than invent.
The workspace-anchored rule supersedes all stylistic conventions. Never name the workspace, the company, or any individual.
Every brief serves four lenses: substantive content, competitive positioning, client-conversation enablement, action.
Format selected by item_type, not by section count target. Brief length is determined by sourced content, not by aspirational length.

Storage Format

Briefs are stored as markdown in intelligence_items.full_brief. The markdown convention:

Each section is a top-level heading (# Section Name)
Section names match the format's section list exactly
Sections that have no grounded content are omitted entirely OR carry a single-line note: "No content for this section as of [date]: [reason]."
Inline citations use the format: Source: [Title], [Issuing Body], [Date]. [URL if applicable].
Severity labels are written as the space-separated form (ACTION REQUIRED, COST ALERT, WINDOW CLOSING, COMPETITIVE EDGE, MONITORING)
Cause-and-effect chains use bullet structure: cause sentence, mechanical-consequence sentence, effect-by-vertical sentences
The workspace is referenced as "the workspace" or "workspaces in [role]" or by operational profile, never by name

This convention enables consistent display in the UI and enables a future schema migration to extract structured fields from the markdown reliably.

Database Field Emission (YAML frontmatter contract)

Every regeneration writes the 19-field contract to intelligence_items (was 13; the metadata-persist work added trajectory_points, what_it_changes, does_not_resolve, conversion_trigger, cross_references, signal_band, theme and the theme_candidate capture-not-null companion). The full_brief column carries the markdown body produced under the format selected above; the metadata fields are emitted as a YAML frontmatter block at the very end of the markdown output, after any New Sources Identified section. Downstream code (synthesiseAndWriteBrief in src/lib/agent/canonical-pipeline.ts, the single write site) parses the YAML, maps each CHECK-constrained field to its live DB vocabulary (src/lib/agent/metadata-vocab.ts), and writes the row; an absent/malformed YAML block, or a metadata write rejected by a constraint, is a failed regeneration (fail-loud, never a silent partial write).

Fields:

full_brief — the markdown body of the brief, structured per the format type's section list. Already produced as the body of the agent's output.
severity — one of the 5 severity labels. Reflects the urgency of action implied by the brief's content as it actually exists, not as it would exist if all sections were filled. Briefs that honestly omit sections under the integrity rule still emit severity, scoped to what is known and sourced.
priority — the 4-tier dashboard counter value, computed from severity per the locked mapping below. The agent computes this; downstream code does not.
urgency_tier — the dashboard tier value, one of watch, elevated, stable, informational.
format_type — the format used for this brief, derived from item_type per the locked mapping below.
topic_tags — array of 0-3 values from the 7 Topic Categories controlled vocabulary above. Reflects what the brief actually covers, not what it nominally is named after. Emitted as a YAML inline array. Empty array allowed when the item genuinely fits none of the seven (rare). Tags outside the vocabulary fail the regeneration.
operational_scenario_tags — array of 0-5 values describing operational scenarios the item touches. Prefer the core glossary in the Operational Scenario Tags section above; new values allowed when the core doesn't fit. Lower-case kebab-case. Drives intersection detection.
compliance_object_tags — array of 0-4 values from the closed Compliance Object Tags glossary above. Tags outside the glossary fail the regeneration. Drives intersection detection.
related_items — UUID array of intelligence_items the agent recognised as related during composition. UUIDs must come from the source pool input. No invented UUIDs. Empty array when no relations identified.
intersection_summary — short markdown string (≤1500 chars) describing how this item interacts with the linked items. Sourced; cite linked items inline by title. Empty string OR null when no intersections were identified.
sources_used — UUID array of source IDs the agent referenced. Populated only with IDs that arrived in the input context. No invented UUIDs.
last_regenerated_at — ISO 8601 timestamp at the moment of generation. Current UTC timestamp in ISO 8601 form (e.g., 2026-04-28T18:42:00Z). Never NOW(), never a placeholder, never derived from source publication dates.
regeneration_skill_version — fixed string identifying the SKILL.md contract version. For regenerations under the current contract, the value is "2026-04-29".

Severity to priority mapping (locked):

ACTION REQUIRED → CRITICAL
COST ALERT → HIGH
WINDOW CLOSING → HIGH
COMPETITIVE EDGE → MODERATE
MONITORING → LOW

format_type derivation from item_type (locked):

regulation, directive, standard, guidance, framework → regulatory_fact_document
technology, innovation, tool → technology_profile
regional_data → operations_profile
market_signal, initiative → market_signal_brief
research_finding → research_summary

Emission format. The agent appends the YAML frontmatter block at the very end of the markdown output, after any New Sources Identified section, fenced with --- delimiters. The block is NOT wrapped in markdown code fences (no triple-backtick yaml). The block stands alone with its --- delimiters as the only fences. Example:

---
severity: ACTION REQUIRED
priority: CRITICAL
urgency_tier: watch
format_type: regulatory_fact_document
topic_tags: [emissions, reporting]
operational_scenario_tags: [CBAM-declaration, customs-declaration-import, emissions-reporting-Scope3]
compliance_object_tags: [importer, customs-broker, manufacturer-producer]
related_items: [b3c4d5e6-f7a8-4901-2345-678901234567]
intersection_summary: "Overlaps with EU ETS for shipping (linked) on emissions-reporting-Scope3; CBAM declarants importing covered goods that arrived via EU-ETS-priced ocean freight face dual reporting obligations on the same emission units."
sources_used: [a1b2c3d4-e5f6-4789-9abc-def012345678, fedcba98-7654-4321-0fed-cba987654321]
last_regenerated_at: 2026-04-29T18:42:00Z
regeneration_skill_version: "2026-04-29"
---

The metadata block is mandatory on every regeneration. An absent or malformed block is a failed regeneration.

Update Protocol

When the user says "update the skill," the agent:

Web searches all sources in the priority registry above
Identifies what changed since the last update
For each change: specifies what changed, previous state, new state, severity label, affected transport modes and cargo verticals, cost impact range
Applies cause-and-effect chain to every finding
Adds any new credible sources discovered during search
Flags any source conflicts as disputes
Generates an updated version of this SKILL.md file with current intelligence, expanded source list, and dated changelog
Delivers as a downloadable file for upload to /mnt/skills/user/environmental-policy-and-innovation/

Changelog

2026-06-05: Section-count consistency. The Regulatory Fact Document was labeled "14 sections" in the format header and the integrity-rule examples while the section bodies define Sections 1–15 (Section 15 Sources). The canonical count is 15 (14 content sections 1–14 + Section 15 Sources); the "14" labels were stale (they counted only content sections). Reconciled all count references to 15, and corrected the worked example from "8 of 14 sections (1,2,3,4,8,10,11,14,15)" — which listed 9 numbers — to "9 of the 15 sections." No section was added or removed; only the count labels and the one miscount were fixed.

2026-04-29: Phase B.2.5 — intersection-readiness contract. Extended the YAML emission contract from 9 fields to 13 fields with four new intersection-readiness fields: operational_scenario_tags (open vocabulary, 0-5 tags, ~36-value core glossary across ocean / air / road / customs-trade / carbon-ETS / reporting / packaging-products), compliance_object_tags (closed vocabulary, 0-4 tags, 18 supply-chain roles spanning carriers / vehicle operators / forwarders-intermediaries / cargo principals / infrastructure operators), related_items (UUID array of intelligence_items the agent recognised as related during composition; integrity rule applies — no invented UUIDs), and intersection_summary (≤1500 chars markdown describing how this item interacts with linked items). Added the Intersection Detection section documenting the system feature: pairs sharing ≥1 operational_scenario_tag AND ≥1 compliance_object_tag are detected automatically, ranked by strength score (+3/scenario, +2/compliance object, +5 for explicit related_items linkage, +2 if both items priority CRITICAL/HIGH), canonicalized so each pair appears once. Added 7 Topic Categories closed-vocabulary documentation and Operational Scenario Tags open-vocabulary core glossary. Bumped regeneration_skill_version from "2026-04-28" to "2026-04-29".

2026-04-28: Major rewrite. Replaced 10-section intelligence brief with 14-section regulatory fact document for regulation/directive/standard/guidance/framework. Added four format-specific output structures: Technology Profile (8 sections), Operations Profile (8 sections), Market Signal Brief (8 sections), Research Summary (6 sections). Added the Integrity Rule (no invented content, sections omitted when ungrounded). Added the Workspace-Anchored Rule (no company names, no personal names, generic language driven by workspace profile). Added Anticipated Authoritative Guidance and Pending Regulatory Events as a structured section in the regulatory fact document. Added the Cross-Format Lens Requirement (every brief serves substantive, competitive, client-conversation, and action lenses). Reframed Technology Profile from technology reference to multi-lens intelligence covering all sustainability technology, not only competitor-access scenarios. Reframed Operations Profile from regional regulatory reference to cost-and-feasibility decision support. Added operator and competitive intelligence sources to priority registry. Standardized severity labels to space-separated form. Established markdown storage convention for full_brief column.

2026-04-11: Unified skill file with agent prompt framework. Added 10-section intelligence brief structure, mandatory cause-and-effect requirement per data point, severity labels (ACTION REQUIRED, COST ALERT, WINDOW CLOSING, COMPETITIVE EDGE, MONITORING), business evaluation framework (core lens: what do I know before competitors and what do I do with that lead time), and context column requirement for all data tables.

2026-03-02: Restructured skill file to match FSI app architecture. 119 resources across 7 categories, 7 topics, 8 jurisdictions. Full impact scoring and urgency calculation methodology documented.