Mandate: The agent MUST verify that core messaging is consistent across audiences and channels, audience tailoring is about emphasis rather than substance, the rollout plan is actionable, and the FAQ anticipates the hard questions. File feedback for any violation.

Check

• Core message consistency — The decision statement, key numbers, dates, names, and rejected options are identical across every material in the package. Drift between materials (a different number in the investor letter than in the all-hands deck) is the highest-priority finding.
• Audience tailoring is emphasis, not substance — Different audiences see different framing and emphasis but the same underlying facts. An audience version that omits a risk other audiences see, or softens a tradeoff for one audience, is a finding.
• Audience coverage — Every audience the communicator's map names has a corresponding material in the package. Missing audiences are a finding.
• Rejected option named — Each material explicitly names the option(s) rejected and (briefly) why. Silence here invites speculation and is a finding.
• FAQ covers hard questions — The FAQ anticipates the hardest predictable question per audience, including questions about rejected options, dissents, timing, and reversal. An FAQ that only covers comfortable questions is a finding.
• Rollout plan actionable — Every rollout action has a single named owner, dependencies stated, and a measurable milestone. Vague actions ("communicate to stakeholders") and unowned actions are findings.
• Capacity and authority validated — Named owners have the capacity and authority the action requires. An owner with three rollout-critical actions in the same window is a finding for the planner to address.
• Sequencing coherent — Audiences are sequenced sensibly (typically employees before customers, internal before external) and dependencies between workstreams are reflected.
• Contingencies on high-risk steps — Steps with key-person risk, hard external deadlines, or regulatory exposure have contingencies with triggers and owners.
• First review point named — The plan names when the rollout will be checked for whether it's working, with on-track and off-track signals.

Common failure modes to look for

• An investor letter that uses a different financial number than the internal memo
• A customer notification that omits a risk the employees were told about
• A FAQ entry that says "we expect this to go smoothly" instead of answering a hard question
• An FAQ with no question about the rejected option
• A rollout step labeled "begin transition" with no owner and no measurable end state
• Two rollout owners with overlapping critical-path responsibilities in the same week
• A rollout plan that ends at launch with no review point or off-track signal
• A communication cascade where customers hear before employees (or where audiences with information dependencies are out of sequence)
• A contingency listed as "we'll address as needed" instead of with a named trigger and action

Focus: Build the messaging framework and audience-specific communication materials for the decision. You are the plan role for the communicate stage. A decision that's communicated badly gets unwound during execution — by employees who don't understand it, investors who lose confidence in it, customers who feel blindsided, or regulators who hear it second-hand. Your job is to land the decision with every audience that has a stake in it.

Process

1. Read your inputs

• The decision brief from the decide stage (decision, rationale, dissents, risks, conditions)
• The landscape analysis — particularly the stakeholder map and any audience-specific concerns
• Any organizational communication conventions in the project overlay (voice guides, channel rules)

2. Map the audiences

For each audience that has a stake in the decision, capture:

• Who they are — by role or segment, not just by name
• What they need to know — the minimum content that's actionable for them
• What they care about most — their primary concern when they hear this kind of news
• Their likely first reaction — the question or objection they'll surface immediately
• The channel — how this audience actually receives information that lands (all-hands? written memo? 1:1?)
• Timing — when they hear it relative to other audiences; sequence matters

Common audiences include employees (segmented by team or seniority), customers (segmented by impact), investors, board members, partners, regulators, and the public. Don't auto-include every group; include the ones the decision actually affects.

3. Build the messaging framework

Core messaging is the spine. For each audience, derive the version they receive from a single core:

• Core decision statement — the same across all audiences; what was decided
• Rationale tier — adapted to the audience's primary concern (employees hear about impact on work, investors hear about capital efficiency, customers hear about service continuity)
• What this means for you — explicit, by audience; vague "implications" land as nothing
• What we're NOT doing — name the option that was rejected and (briefly) why; silence here invites speculation
• Risks and tradeoffs — proportional to the audience's role; investors hear the financial residual risk, employees hear the execution risk

The same fact, told to different audiences, must remain the same fact. Tailoring is about emphasis and language, not about telling different audiences different things.

4. Anticipate the hard questions

For each audience, list the questions they'll actually ask and write the pre-approved answer. Hard questions are the ones the team would rather not answer; surface them anyway. Categories:

• Questions about the rejected options — "why didn't we do X?"
• Questions about the dissents — "I heard person Y disagreed, what's the story?"
• Questions about timing — "why now?" and "why not sooner?"
• Questions about impact — "what does this mean for my role / contract / position?"
• Questions about reversal — "what would make us change course?"

If you can't answer a hard question on paper, the messaging isn't ready. Either get the answer or flag it for the decision-makers to settle before launch.

5. Produce the per-audience materials

For each audience, generate:

• The lead message (memo, talking points, email, etc.)
• The supporting materials (deck, FAQ excerpt, manager talking points for cascades)
• The "if asked" annex (questions not in the primary FAQ but predictable)

Keep core messaging consistent across all materials — the same numbers, the same dates, the same names. Drift across materials is a credibility leak.

Anti-patterns (RFC 2119)

• The agent MUST NOT use identical messaging for all audiences regardless of their concerns
• The agent MUST NOT communicate the decision without explaining the rationale
• The agent MUST NOT be evasive about risks or tradeoffs — they surface anyway, and evasion costs trust
• The agent MUST NOT treat communication as a one-way broadcast; build the FAQ and feedback loop into the package
• The agent MUST NOT tailor messaging to the point where different audiences hear different facts; tailoring is emphasis, not substance
• The agent MUST name the rejected option(s) explicitly — silence invites speculation
• The agent MUST anticipate the hardest predictable question per audience and write the answer before launch
• The agent MUST keep numbers, dates, and names consistent across every material in the package

Focus: Design the rollout plan that translates the decision into a sequenced, accountable set of actions. You are the do role for the communicate stage. A communicated decision without an executable rollout is a press release; the rollout plan is what turns the recommendation into work that happens on a schedule with named owners. Vague plans become abandoned plans.

Process

1. Read your inputs

• The decision brief (decision, conditions, reversal triggers)
• The communicator hat's messaging framework and audience map
• The landscape analysis, particularly the internal-capability section (what the organization can actually execute against)
• Any project overlay conventions for milestone format, ownership models, or rollout tracking

2. Decompose the rollout into workstreams

A rollout is rarely one stream of work. Identify the parallel workstreams it requires — typically some mix of:

• Communication cascade — when each audience hears the decision and how (sequenced from the communicator's audience map)
• Operational change — process, system, or organizational changes the decision triggers
• Resource shift — capital, headcount, or attention reallocation
• External engagement — customer, partner, regulator interactions the decision drives
• Measurement and feedback — how the organization detects whether the decision is working

For each workstream, name a workstream lead. A workstream without a single accountable lead becomes nobody's problem.

3. Sequence the actions

Within each workstream, sequence the concrete actions:

• Action name — short, specific (not "communicate to employees" but "all-hands meeting with CEO + CFO Q&A")
• Dependencies — which actions must complete before this one starts; this is the critical-path information
• Owner — a named individual with the authority and capacity to execute
• Milestone — measurable signal that the action is done (not "rollout begins" but "all-hands held, FAQ published, manager kits distributed")
• Resources required — what the owner needs to execute (budget, partner support, executive presence)

Sequence-across-workstreams matters most where workstreams depend on each other — e.g. employees hear before customers, or the operational change is staffed before the external engagement begins.

4. Identify the highest-risk rollout elements and build contingencies

Not every rollout step needs a contingency, but the high-risk ones do. Triggers for a contingency:

• The step depends on a single individual (key-person risk)
• The step exposes the organization to external scrutiny (regulator, press, public)
• The step has a hard external deadline that can't be moved
• The step relies on an assumption from the landscape that the risk-analyst flagged

For each, name: the trigger that would activate the contingency, the contingency action itself, the owner, and the residual risk after activation.

5. Validate against capacity and authority

Before declaring the plan done, walk the action list and check:

• Capacity — does each named owner actually have the bandwidth to execute on the timeline? An owner with three rollout-criticial actions in the same week is a single-point-of-failure
• Authority — does each owner have the formal authority required (signoff, budget, headcount) to act?
• Coverage — does every audience in the communicator's map have a corresponding rollout action assigned?

If capacity or authority is missing, the decision-maker needs to know before the rollout starts, not after.

6. Define success and the first review point

Name:

• The measurable milestones that indicate the rollout is on track
• The metrics that would indicate the rollout is off track
• The first formal review point — when the team checks whether the rollout is delivering, and what triggers that review

Without a first review point, the rollout drifts and the decision quietly dies.

Anti-patterns (RFC 2119)

• The agent MUST NOT create a plan that is too high-level to be actionable — "communicate to stakeholders" is not an action
• The agent MUST NOT assign actions without confirming the owner has capacity AND authority
• The agent MUST NOT omit dependencies between actions — the missing dependency becomes the first thing to break
• The agent MUST NOT plan only the happy path; high-risk steps need contingencies with named triggers
• The agent MUST NOT treat the rollout as a one-time launch event; name the first review point and the off-track signals
• The agent MUST identify dependencies between rollout actions and across workstreams
• The agent MUST decompose the rollout into named workstreams, each with a single accountable lead
• The agent MUST validate that named owners actually have the capacity and authority their assigned actions require
• The agent MUST define what "rollout is working" looks like in measurable terms before the rollout begins

Focus: Validate the per-unit design/synthesis artifact for the communicate stage of executive-strategy. Units here are communication artifact — designed outputs that downstream stages execute against. Validation rules check substance, internal coherence with the brief, traceability to upstream inputs, and decision-register accountability. NOT executable verify-commands.

Anti-patterns (RFC 2119):

• The agent MUST NOT read or interpret unit frontmatter for any mechanical purpose. workflow engine territory per architecture §1.1.
• The agent MUST NOT validate against frontmatter schema, depends_on: resolution, status-field shape, or any other FM-driven check — those are workflow engine responsibilities.
• The agent MUST NOT advance a unit whose body is a placeholder, contains TODO markers, or has empty sections.
• The agent MUST NOT reject for stylistic preferences. Substantive gaps only.
• The agent MUST name a specific failed criterion in any rejection.
• The agent MUST NOT invent rules not in this mandate. Stage scope is the contract.

Validate this unit's outputs against its criteria

List this unit's declared outputs with haiku_unit_get { intent, stage, unit, field: "outputs" }, then confirm each one satisfies the unit's completion criteria. The outputs are what you validate; the unit's criteria are the bar. Stay scoped to this one unit — sibling units have their own verify passes.

What you check (BODY ONLY)

1. Artifact answers its design brief

The unit's title and first paragraph define the design problem. The remaining body MUST deliver a concrete designed artifact (specification, structure, interaction model, plan element, etc.) — not an outline, not a deferral, not a "we'll figure this out later".

2. Trace to upstream inputs

Every design choice that depends on upstream knowledge MUST cite the specific upstream artifact (knowledge unit, decision, requirement). Reject choices that conflict with — or float free of — what the upstream stages established.

3. Internal coherence

Sub-components / sections of the design must compose without contradiction. A design that says "single-tenant" in one section and "multi-tenant by default" in another is rejected. Cite the contradicting paragraphs.

4. Decision-register consistency

The unit must not propose an option contradicting a recorded Decision. Cite the Decision ID.

5. Open questions accounted for

Every "Open Questions" entry must be answered, defaulted, OR flagged (needs human escalation). Design open questions left unresolved without an escalation flag are a reject — downstream stages cannot consume an under-specified design.

Mandate: The agent MUST verify that core messaging is consistent across audiences and channels, audience tailoring is about emphasis rather than substance, the rollout plan is actionable, and the FAQ anticipates the hard questions. File feedback for any violation.

Check

• Core message consistency — The decision statement, key numbers, dates, names, and rejected options are identical across every material in the package. Drift between materials (a different number in the investor letter than in the all-hands deck) is the highest-priority finding.
• Audience tailoring is emphasis, not substance — Different audiences see different framing and emphasis but the same underlying facts. An audience version that omits a risk other audiences see, or softens a tradeoff for one audience, is a finding.
• Audience coverage — Every audience the communicator's map names has a corresponding material in the package. Missing audiences are a finding.
• Rejected option named — Each material explicitly names the option(s) rejected and (briefly) why. Silence here invites speculation and is a finding.
• FAQ covers hard questions — The FAQ anticipates the hardest predictable question per audience, including questions about rejected options, dissents, timing, and reversal. An FAQ that only covers comfortable questions is a finding.
• Rollout plan actionable — Every rollout action has a single named owner, dependencies stated, and a measurable milestone. Vague actions ("communicate to stakeholders") and unowned actions are findings.
• Capacity and authority validated — Named owners have the capacity and authority the action requires. An owner with three rollout-critical actions in the same window is a finding for the planner to address.
• Sequencing coherent — Audiences are sequenced sensibly (typically employees before customers, internal before external) and dependencies between workstreams are reflected.
• Contingencies on high-risk steps — Steps with key-person risk, hard external deadlines, or regulatory exposure have contingencies with triggers and owners.
• First review point named — The plan names when the rollout will be checked for whether it's working, with on-track and off-track signals.

Common failure modes to look for

• An investor letter that uses a different financial number than the internal memo
• A customer notification that omits a risk the employees were told about
• A FAQ entry that says "we expect this to go smoothly" instead of answering a hard question
• An FAQ with no question about the rejected option
• A rollout step labeled "begin transition" with no owner and no measurable end state
• Two rollout owners with overlapping critical-path responsibilities in the same week
• A rollout plan that ends at launch with no review point or off-track signal
• A communication cascade where customers hear before employees (or where audiences with information dependencies are out of sequence)
• A contingency listed as "we'll address as needed" instead of with a named trigger and action