🎯 Output Router

The same insight, restructured for the channel that will actually be read — not just reformatted, but rebuilt for how your audience processes information.

Core Philosophy

Every format has a reading contract — an implicit agreement between the content and the reader about how the information will be structured, what the reader needs to bring, and how much time it demands. A five-page analysis report makes a reading contract that says: bring 20 minutes, follow the argument linearly, tolerate uncertainty as the logic builds. An executive summary makes a contract that says: give me 90 seconds and I will tell you the three things that matter. A slide deck makes a contract that says: each frame is one idea, and the logic connects across frames rather than within them.

Transforming content between formats is not a cosmetic operation. Cutting an analysis report down to a five-sentence summary is not the same as transforming it. True transformation requires identifying the structural logic of the source, then rebuilding that logic according to the target format's conventions. A good executive summary does not simply remove words from the analysis; it answers the question "what does the decision-maker need to act?" — which is a different question than the analysis was built to answer. A good slide deck does not extract sentences and put them on slides; it redesigns the argument as a series of single claims, each defensible on its own, that add up to a conclusion.

This skill also exists to prevent a common failure: over-engineering the format for the audience. A team update that goes out as a 12-slide deck when an email would do is not more rigorous — it is harder to read, less likely to be consumed, and signals poor judgment about the audience's time. The output-router includes a fit-check step precisely to guard against this: the goal is the most effective format for this audience and this purpose, not the most elaborate.

🏗️ ANCT Architecture

Entropy Profile:

📥 Intake Variables

{{SOURCE_MATERIAL}}

{{CHANNEL}}

{{AUDIENCE}}

{{CONSTRAINTS}}

{{OBJECTIVE}}

Supported Transformations

⚙️ Phases

Phase 1: Intake Content and Target

Entropy Level: E1 – Deterministic Control Mode: DELEGATE

Collect the source material and transformation parameters before any analysis begins. The most critical intake field is

{{OBJECTIVE}}

Also probe for

{{CONSTRAINTS}}

Process:

1. Collect

   {{SOURCE_MATERIAL}}

   ,

   {{CHANNEL}}

   ,

   {{AUDIENCE}}

   , and

   {{OBJECTIVE}}

2. Identify the source format by examining the source material structure (narrative vs. structured vs. data-heavy vs. mixed)
3. Confirm the target format is in the supported transformations list, or ask for clarification if the request is non-standard
4. Probe for constraints: length limit? named style guide? any must-include or must-exclude elements?
5. Assess source material completeness: if the source is fragmentary, ambiguous, or clearly a draft, flag this — transformation of incomplete source produces incomplete output

Output: A confirmed intake record with source format identified, target format confirmed, audience defined, objective stated, and constraints captured.

Phase 2: Analyze Content Structure

Entropy Level: E3 – Analytical Control Mode: NARRATE

Before transforming, understand what you have. Identify the source material's structural elements: key claims (the assertions the content is making), supporting evidence (the data or reasoning behind each claim), action items (specific asks or decisions), and context-setting (background that is needed to understand the claims but not the claims themselves).

This structural analysis drives the transformation decision: which elements must survive the transformation intact, which can be compressed, which can be merged, and which can be dropped. Action items always survive intact. Key claims survive, though they may be restructured. Supporting evidence is compressible — the level of detail kept depends on the target format and audience need. Context-setting is usually the most aggressively compressed element in summary transformations.

Process:

1. Read the entire source material before beginning analysis — do not analyze paragraph by paragraph
2. Extract and list all key claims (the things the content is asserting as true or important)
3. Extract and list all action items (explicit asks, decisions, owners, deadlines)
4. Identify the primary conclusion or so-what — the single most important thing the audience needs to know
5. Map supporting evidence to each key claim — note which claims are well-supported vs. asserted without evidence
6. Classify each structural element: must-preserve / can-compress / can-drop for the target format
7. Flag any gaps: claims in the source that lack supporting evidence, or questions the source raises but does not answer

Output: A content structure map showing key claims, action items, primary conclusion, evidence mapping, and a must-preserve / can-compress / can-drop classification for target format transformation.

Phase 3: Transform and Adapt

Entropy Level: E4 – Generative then E3 – Analytical Control Mode: GENERATE then NARRATE

Apply the transformation. GENERATE executes the format change using the structural map from Phase 2. NARRATE then validates completeness: every must-preserve element from Phase 2 must appear in the output; no action items may be lost; the primary conclusion must be present.

The most important principle in this phase: apply the target format's structural conventions, not just its visual conventions. A bullet list is not just a paragraph with hyphens added — it is a format that assumes parallel structure, equivalent level of abstraction across items, and an implicit claim that each item stands alone. A slide deck is not a report with images — it is a format where each slide must make one claim that is defensible on its own. An executive summary is not a short report — it is a format that answers "what should I know and what should I do?" before answering "why."

Process:

1. Apply the target format transformation using the content structure map as input
2. For each target format, apply its specific structural conventions:
   • Executive summary: BLUF (bottom line up front), maximum 5 sentences, no methodology, one recommendation
   • Slide deck: one claim per slide, visual logic (not just text), explicit "so what" on final slide
   • Structured table: consistent column headers, parallel row structure, counts where applicable
   • Plain language: active voice, no jargon, one idea per sentence, analogy for any technical concept
   • 1-page brief: inverted pyramid (finding → evidence → background), hard page constraint enforced
   • Action item register: owner / action / deadline format, no discussion context
3. NARRATE validation: cross-check output against the must-preserve list; flag any missing elements
4. Check that the transformed output answers

   {{OBJECTIVE}}

   — not just that it is in the right format

Output: A transformed version of the source content in the target format, with a completeness check confirming all must-preserve elements are present.

Phase 4: Validate Fit

Entropy Level: E3 – Analytical Control Mode: NARRATE

Check that the output is genuinely optimized for the target format — not just reformatted but properly restructured. There is a common failure mode in format transformation where the output looks like the target format but retains the logical structure of the source format. A bullet list that is actually a series of sentence fragments from a paragraph is not transformed; it is truncated. A slide that contains three paragraphs of text is not a slide; it is a report displayed on a slide.

Also check fit for audience: does the level of detail, vocabulary, and assumed context match

{{AUDIENCE}}

Process:

1. Apply format-specific fit criteria to the output:
   • Executive summary: Does it answer "so what?" in the first sentence? Is it under 5 sentences? Does it contain any methodology or background?
   • Slide deck: Does each slide have exactly one claim? Are there any slides with more than 40 words of body text? Is there an explicit "so what" conclusion slide?
   • Bullet list: Are all bullets parallel in structure? Are any bullets actually mini-paragraphs?
   • Table: Are all columns meaningful? Are rows internally consistent? Are counts or frequencies included where useful?
   • Plain language: Is the Flesch-Kincaid reading level appropriate? Are all technical terms either explained or removed?
2. Check audience fit: vocabulary, assumed knowledge, level of detail
3. Check length fit against

   {{CONSTRAINTS}}

   — if a word or slide limit was specified, enforce it
4. Produce a fit assessment: pass / needs revision for each criterion, with specific revision recommendations for any failures

Output: A fit validation report with pass/fail assessment against format criteria and audience fit, plus specific revision recommendations.

Phase 5: Produce Formatted Output

Entropy Level: E1 – Deterministic Control Mode: DELEGATE

Deliver the final transformed output in clean, channel-ready format. If the target is a document, apply appropriate heading structure, white space, and visual hierarchy. If the target is a slide deck, deliver the slide structure with title, body claim, and speaker notes for each slide. If the target is a table, deliver it with headers, sorted by frequency or priority, with a summary row where appropriate.

Include a source-to-output mapping note if the requester might need to audit the transformation — particularly useful when the source is long and the output is highly compressed, and the requester needs to verify that key claims survived intact.

Process:

1. Apply final formatting to the validated output from Phase 4
2. For slide deck outputs: deliver each slide as Title / Body Claim / Speaker Notes structure
3. For table outputs: sort by most meaningful dimension (frequency, priority, or chronology), include totals or summary rows, apply consistent column widths
4. For document outputs: apply heading hierarchy, ensure consistent paragraph length, add metadata (date, audience, source document reference)
5. If source-to-output mapping was requested: produce a side-by-side mapping showing which source element each output element corresponds to
6. Add a brief transformation note: what was preserved, what was compressed, what was dropped, and why

Output: Final formatted output in the target format, channel-ready, with optional source-to-output mapping and transformation notes.

🚫 Anti-Patterns

💡 Example Applications

🖥️ Platform Notes

CLI: Ideal for piping source documents into the transformation workflow and outputting structured formats (markdown, CSV for tables, JSON for visualization specs). Useful in automated content pipelines where report outputs need to be transformed for multiple audiences simultaneously.

Web: Best for interactive refinement. Generate the transformation, then iterate on specific sections. The fit validation in Phase 4 is particularly useful as an interactive conversation where the requester can confirm which must-preserve elements matter most.

IDE: Useful for teams maintaining content templates. The transformation logic can be embedded in documentation workflows where the same analysis automatically produces multiple output variants (report, summary, deck) from a single source document.

Any LLM: Haiku is appropriate for most straightforward transformations (bullets to narrative, narrative to bullets, table formatting). For transformations requiring strong structural reasoning (executive summary compression with claim preservation, slide deck logic design), sonnet is recommended.

📋 Compliance

AI Governance Alignment: Output transformation is a low-risk operation from a governance perspective. The key compliance consideration is fidelity: the transformation must not materially alter the meaning of source claims. The completeness validation in Phase 3 and the fit validation in Phase 4 are the primary controls for transformation fidelity.

PII Risk Level: low — The output-router transforms content structure, not content substance. However, if

{{SOURCE_MATERIAL}}

Model Recommendation: haiku — for most format transformations (bullet ↔ narrative, table formatting, plain language rewrites). Upgrade to sonnet when the transformation requires strong claim-extraction reasoning (executive summary compression, slide deck logic design, or any transformation where the source is analytically complex).

Data Handling: No data is retained between sessions. If source material contains sensitive or confidential content, ensure the output format is appropriate for the sensitivity level of the audience — a highly compressed summary may be appropriate for a broader audience than the full source document.