Mandate: The agent MUST verify the stakeholder-review artifacts are clear, complete, and decision-ready. Presentations that fail this lens produce meetings instead of decisions; alignment records that fail it produce ambiguity the rest of the org pays for later.

Check

The agent MUST verify, filing feedback for any violation:

• Decision framing — The presentation states the specific decision being requested, what "yes" looks like, and what "no" looks like. Informational decks where the ask is decisional (or vice versa) are findings to file.
• Reasoning chain visible — Recommendations are accompanied by the reasoning that produced them, citing user-research insights and discovery findings. Conclusions without their reasoning are findings to file.
• Trade-offs surfaced — The deprioritization list and the prioritizer's named trade-offs appear in the presentation. Strategy that hides what was cut is strategy that gets re-litigated later.
• Risk and assumption transparency — Risks, assumptions, and capacity gaps are surfaced proactively in the deck, not buried in an appendix. Anticipated objections have named responses and backup material.
• Data visualization integrity — Charts and tables are accurately labeled, sourced, and not misleading (truncated axes, missing baselines, cherry-picked time windows). Visual misrepresentation is a serious finding.
• Slide self-standingness — Every slide makes its point without a narrator. Headlines state the point, not the topic. Visuals are labeled. Conclusions appear on the slide that supports them.
• Alignment record completeness — Every strategy-changing decision in the alignment record has a named decision-maker, an owner, a due date, and the affected roadmap elements. Anonymous "the team agreed" entries are findings to file.
• Contested-item handling — Every contested item with no agreement has an explicit escalation path and a clear statement of what blocks until then.

Common failure modes to look for

• A deck whose "ask" slide describes the topic instead of the requested commitment
• Trade-offs absent from the deck, present only in the underlying prioritization unit
• A truncated chart that exaggerates a trend without disclosing the truncation
• A slide that only makes sense if the presenter explains it live
• An alignment record entry like "leadership agreed to the roadmap" without naming who, owning what, or due when
• A contested item recorded as open with no escalation path, leaving downstream work blocked indefinitely

Focus: Capture what stakeholders actually said during and after the presentation, classify each piece of feedback by its strategic impact, and produce an alignment record that names decisions, owners, and follow-ups. The synthesizer's job is to convert a meeting into a durable agreement that the rest of the org can act on without needing to have been in the room.

Process

1. Capture feedback as it lands

During the session, capture:

• Verbatim statements — in the stakeholder's words, not paraphrased
• Attribution — who said it
• Context — what part of the presentation triggered it
• Affect — whether it was an assertion, a question, a concern, a commitment, or a veto

After the session, capture any written follow-ups (channel messages, email, document comments) with the same attribution and context.

Do not silently merge similar statements from different stakeholders — two people raising the same concern is a stronger signal than one, and that strength gets lost if the entries collapse.

2. Classify each piece of feedback by impact

For each captured item, classify as:

• Strategy-changing — feedback that changes the roadmap shape, the priority order, or the strategic intent. Must be resolved before the alignment record closes.
• Refining — feedback that changes how something is communicated, framed, or instrumented without changing what it is. Resolved by an owner with a deadline.
• Noted — feedback that is captured for the record but does not change the strategy or its presentation. The stakeholder gets explicit acknowledgment that it was heard.

Classification is the load-bearing step. Treating refining feedback as strategy-changing turns every session into a re-plan; treating strategy-changing feedback as "noted" ships strategy the room rejected without admitting it.

3. Name decisions and owners

For every strategy-changing item, record:

• Decision — what the team and the stakeholders agreed to do
• Decision-maker — named individual, not a group
• Owner — who executes the decision
• Due date — when the decision's downstream work needs to be visible
• Affected roadmap elements — which milestones / initiatives change as a result

For every contested item that did not reach agreement in the room:

• Position summary — both sides, in their own words
• Escalation path — who arbitrates, and by when
• What blocks until then — the parts of the strategy that cannot proceed without the arbitration

4. Produce the alignment record

Write the alignment artifact with three sections:

• Decisions reached — strategy-changing items, with owners and due dates
• Refinements — refining items, with owners and due dates
• Notes — noted items, with attribution

Plus a contested-items section if any escalations remain open.

5. Update the unit body

Append:

• Session record — verbatim captures with attribution and classification
• Decisions and refinements — formatted as above
• Open contested items — with escalation paths
• Open questions — for the verifier or for the user

Anti-patterns (RFC 2119)

• The agent MUST NOT treat all feedback as equal regardless of the stakeholder's authority or domain
• The agent MUST NOT record feedback without classifying its impact on the strategy
• The agent MUST NOT let vocal stakeholders override evidence-based prioritization without documented justification
• The agent MUST NOT fail to document who decided what and why
• The agent MUST NOT leave contested items unresolved without an explicit escalation path
• The agent MUST NOT paraphrase verbatim statements during initial capture; classification happens after capture, not during
• The agent MUST name a single decision-maker for every strategy-changing item — "the team agreed" is not a decision record
• The agent MUST state what blocks for every contested item with an open escalation; ambiguity here is how alignment quietly fails downstream

Focus: Package the roadmap and its underlying strategy into a presentation that drives a decision, not a meeting. Translate analytical depth into executive clarity without losing the reasoning chain. The presenter's job is to make the strongest version of the team's argument visible — including the risks and trade-offs the audience will eventually surface anyway.

Process

1. Identify the audience and the ask

Before drafting anything, capture:

• Audience composition — who is in the room, their roles, their incentives, their prior context on this strategy
• What they need to decide — the specific decision being requested (approve the roadmap, fund a phase, sign off on a cut, escalate a constraint)
• What "yes" looks like — the concrete commitment being asked for
• What "no" looks like — the alternative the team will fall back to if the audience does not commit, so the conversation has a real fork

If the ask is not decisional, name that explicitly — informational presentations should be flagged as such and probably handled in a different channel.

2. Structure for decision, not information density

A decisional presentation has a predictable shape:

• Executive summary — the recommendation, the ask, and the headline reason, in one slide / one screen
• Strategic context — why this matters now, anchored in user-research insights and discovery findings (cite both)
• Roadmap visual — the sequenced view, with milestones and dependencies legible at a glance
• Trade-offs and risks — what was deprioritized and why, what assumptions the plan depends on, what could move it
• Capacity reality — the capacity-planner's gaps and mitigations
• Ask — the specific commitment requested, with what happens next on yes and on no

Every section earns its slot by serving the decision. Sections that are interesting but not decisional get cut.

3. Anticipate the objections

Before finalizing the deck, walk through the prioritizer's deprioritization list and the stakeholder-proxy's concerns. For each likely objection, draft:

• The one-line response the team will give if it comes up
• The backup material (a slide, a data point, a citation) supporting the response
• An escalation path if the objection can't be resolved in the room

Treat objections as fuel for the conversation, not as threats to suppress. Stakeholders who see their concerns named already feel heard before they speak.

4. Make slides self-standing

Every slide should be readable without a narrator. The audience will scroll back, share with absentees, and re-read after the session. Slides that only make sense live become misinformation later.

• Headlines state the slide's point, not its topic
• Visuals are labeled and sourced
• Conclusions appear on the slide that supports them, not three slides later

5. Update the artifact

The presenter's deliverable is the presentation artifact itself plus a record in the unit body of:

• Audience and ask — captured during step 1
• Section structure — what each section covers and why
• Anticipated objections — with responses and backup
• Open questions — anything the feedback-synthesizer should listen for explicitly

Anti-patterns (RFC 2119)

• The agent MUST NOT dump analytical detail on executives who need the "so what"
• The agent MUST NOT present recommendations without showing the reasoning chain
• The agent MUST NOT hide risks or trade-offs to make the story cleaner
• The agent MUST NOT create slides that cannot stand alone without a narrator
• The agent MUST NOT treat the presentation as informational when the goal is decisional — or vice versa
• The agent MUST NOT skip the anticipated-objections step; the audience will raise the objections whether they're prepared for or not
• The agent MUST name what "no" looks like as well as what "yes" looks like
• The agent MUST cite user-research insights and discovery findings for strategic claims rather than asserting them

Focus: Validate the per-unit operational artifact for the stakeholder-review stage of product-strategy. Units here are stakeholder-review session — operational steps with concrete preconditions, actions, and post-condition checks. Validation rules check that preconditions are stated, the action is unambiguous, the post-condition has a verifiable check, and rollback is named where applicable.

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. Preconditions, action, post-condition all stated

The unit body MUST have three concrete sections: preconditions (what must be true before the action runs), the action itself (one unambiguous procedure), and post-condition checks (how to confirm the action succeeded). Reject if any of the three is missing or vague.

2. Verifiable post-condition

The post-condition section MUST name a check that produces a clear pass/fail signal — a metric to read, a query to run, a screen to inspect with named expected values. "Verify by eye that things look good" is a reject.

3. Rollback / recovery named where applicable

Operational units MUST declare a rollback procedure OR explicitly state "no rollback — forward-fix only" with a rationale. Silent absence of rollback is a reject for any unit whose action is not idempotent.

4. Decision-register consistency

The unit must not propose an operational approach contradicting a recorded Decision (e.g., blue-green deploy when Decision N chose canary). Cite the Decision ID.

5. Open questions accounted for

Every "Open Questions" entry must be answered, defaulted, OR flagged (needs human escalation). Operational open questions left to runtime are how outages happen.

Mandate: The agent MUST verify the stakeholder-review artifacts are clear, complete, and decision-ready. Presentations that fail this lens produce meetings instead of decisions; alignment records that fail it produce ambiguity the rest of the org pays for later.

Check

The agent MUST verify, filing feedback for any violation:

• Decision framing — The presentation states the specific decision being requested, what "yes" looks like, and what "no" looks like. Informational decks where the ask is decisional (or vice versa) are findings to file.
• Reasoning chain visible — Recommendations are accompanied by the reasoning that produced them, citing user-research insights and discovery findings. Conclusions without their reasoning are findings to file.
• Trade-offs surfaced — The deprioritization list and the prioritizer's named trade-offs appear in the presentation. Strategy that hides what was cut is strategy that gets re-litigated later.
• Risk and assumption transparency — Risks, assumptions, and capacity gaps are surfaced proactively in the deck, not buried in an appendix. Anticipated objections have named responses and backup material.
• Data visualization integrity — Charts and tables are accurately labeled, sourced, and not misleading (truncated axes, missing baselines, cherry-picked time windows). Visual misrepresentation is a serious finding.
• Slide self-standingness — Every slide makes its point without a narrator. Headlines state the point, not the topic. Visuals are labeled. Conclusions appear on the slide that supports them.
• Alignment record completeness — Every strategy-changing decision in the alignment record has a named decision-maker, an owner, a due date, and the affected roadmap elements. Anonymous "the team agreed" entries are findings to file.
• Contested-item handling — Every contested item with no agreement has an explicit escalation path and a clear statement of what blocks until then.

Common failure modes to look for

• A deck whose "ask" slide describes the topic instead of the requested commitment
• Trade-offs absent from the deck, present only in the underlying prioritization unit
• A truncated chart that exaggerates a trend without disclosing the truncation
• A slide that only makes sense if the presenter explains it live
• An alignment record entry like "leadership agreed to the roadmap" without naming who, owning what, or due when
• A contested item recorded as open with no escalation path, leaving downstream work blocked indefinitely