@maestroagora/agora 1.9.0 → 1.10.0

package/skills/agora/references/agora-case-study-runtime.md

@@ -0,0 +1,31 @@
+# Agora case-study runtime
+
+Use this compact file for ordinary customer-success, creative-portfolio, and technical-implementation cases. Load the deep case-study reference only for explicit source review, attribution or permission review, confidentiality review, technical evaluation, audit work, or maintainer analysis.
+
+## Choose the case job
+
+- `CUSTOMER_SUCCESS`: the customer's situation, what changed, and the supplied result.
+- `CREATIVE_PORTFOLIO`: the brief, constraints, decisions, role, work, and supplied result.
+- `TECHNICAL_IMPLEMENTATION`: the system constraint, chosen approach, implementation, observed behavior, and tradeoffs.
+
+Keep real, fictional, mock, and concept status as the user supplies it. Do not add a status label unless requested.
+
+Academic and clinical case reports are outside this runtime.
+
+## Preserve what happened
+
+Use only supplied history, roles, facts, quotations, measurements, and outcomes unless invention is authorized. Do not turn chronology into causation. Keep measured results, reported results, observed changes, supported inferences, targets, pending measurements, and unknowns distinct when the brief distinguishes them.
+
+Do not invent customer motives, internal reactions, implementation steps, permissions, confidentiality terms, quotes, or typicality.
+
+## Shape the case
+
+Start with the most useful situation or result for the intended reader. Explain the constraint, the relevant decision, the work, and what changed. Give each section a distinct job. Keep failed attempts or tradeoffs when they explain the decision.
+
+State each limitation once, beside the fact it limits. Do not repeat the same causal or outcome limit in a closing recap.
+
+For a public case, translate internal review language into customer language. For a technical audience, keep exact system terms and behavior. For sales work, do not turn the case into a chronology or an academic report.
+
+## Return the case
+
+Return the requested case only. Keep source review, permission checks, and internal classifications private unless the user asks for them.

package/skills/agora/references/agora-conversion-runtime.md

@@ -0,0 +1,49 @@
+# Agora conversion runtime
+
+Use this compact overlay for landing pages, product pages, pricing pages, paywalls, checkout, forms, onboarding, upgrade paths, and conversion experiments. Use the deep conversion research reference only when the user asks for research-backed guidance, source review, experiment interpretation, an audit, or maintainer analysis.
+
+## Start from the decision
+
+Identify the reader's current state, the action the surface asks them to take, what happens after that action, and the business outcome the task actually names.
+
+Do not confuse a click with activation, retention, revenue, or another later outcome. Do not claim that a pattern always converts better.
+
+## Keep the fact set closed
+
+Use only supplied routes, prices, terms, commitments, permissions, timing, data use, product behavior, and results. Do not infer self-service access, a sales handoff, a trial, a demo, a download, a confirmation screen, or an implementation step that the brief does not provide.
+
+Keep parallel options separate when they differ in access, effort, exclusivity, price, or commitment. Unknown transitions remain unknown.
+
+Classify each requested component as implementation direction, visible copy, or both. Keep editorial directions and worksheet labels outside ready-to-use copy.
+
+## Make the action clear
+
+State the benefit in reader language, then give the minimum supplied detail needed to believe or choose it. Put the strongest relevant supporting fact near the doubt it answers.
+
+The main action must name its object or destination. Do not use a prerequisite as the main action when the desired event happens later. Keep optional post-action steps secondary.
+
+For a multi-step flow, trace the current state, control, immediate effect, next visible state, and any conditional later state. Keep unsupplied states unknown. Optional consent stays separate from the primary action.
+
+## Match effort and commitment
+
+Make field effort, required information, eligibility, timing, price, renewal, cancellation, and reversibility visible where they affect the choice. Do not hide a material commitment in distant copy.
+
+For pricing, make each available path understandable from the supplied facts. Do not invent feature parity, savings, plan eligibility, negotiation, or billing terms.
+
+Preserve exact prices, billing periods, renewal terms, cancellation rights, trial terms, access timing, collection timing, capability names, regions, role names, availability windows, document status, access conditions, and exclusions. Arithmetic can establish a total, not an unsupplied sequence or collection date.
+
+Use quotation marks only for exact wording supplied as a quotation. A summary of what someone said supplies meaning, not exact speech.
+
+For onboarding, distinguish account creation, setup, activation, and later use. Name the event the task wants to improve.
+
+## Handle experiments
+
+When the user asks for an experiment plan, define the changed element, the eligible population, the primary outcome, guardrails, duration or stopping rule when supplied, and the decision the result will inform. Keep the plan proportional to the request.
+
+When the user asks for interpretation, preserve conflicts, null results, uncertainty, and differences between intermediate and business outcomes. Do not turn one test into a universal rule.
+
+After factual and route requirements pass, preserve the strongest supported reader job as the argument. Do not flatten the surface into an internal evaluation checklist.
+
+## Return the surface
+
+Return the requested copy or implementation decision. Do not append an internal review, rationale, or checklist unless requested.

package/skills/agora/references/agora-craft.md

@@ -527,33 +527,25 @@ The strongest surviving machine-writing tell is cadence rather than vocabulary.
 
 ### The rhythm targets
 
-Every value in the table below is a **governance default**: a control setting chosen so the system has something checkable to enforce instead of the instruction to sound less uniform. The research grades all of them as engineering heuristics with no controlled literature establishing universal values for good prose. None of them is a finding, none may be cited as one, and each is a review trigger rather than an error.
+Numeric sentence-length and paragraph-shape measurements belong in `voice check` and optional diagnostics. They do not set drafting targets.
 
-| Control | Default | Status | Reading |
-|---|---|---|---|
-| Sentence-length spread | For expository prose of at least 20 sentences, a standard deviation of 8 to 12 words with a coefficient of variation of 0.40 to 0.70 | Governance default | Below roughly 6 words of deviation at an 18 to 22 word mean, review for metronomic pacing. Above roughly 15, review for uncontrolled extremes. |
-| Length tails | In a passage of 10 or more sentences, include at least one sentence of 8 words or fewer and one of 25 or more, but only where both are natural | Governance default | A forcing mechanism against middle-length clustering. Drop it wherever it produces a stunt sentence. |
-| Consecutive similar lengths | Flag 3 consecutive sentences when every pair differs by 3 words or fewer | Governance default | Review for accidental monotony. Deliberate parallelism overrides the flag. |
-| Paragraph variation | Over 8 or more paragraphs, aim for a coefficient of variation of at least 0.45 in words | Governance default | Prevents identical blocks without demanding arbitrary one-sentence paragraphs. |
-| Opening structure | At most 2 consecutive sentences opening with the same class, such as subject-first declarative, initial subordinate clause, or imperative | Governance default | Catches repeated `Subject plus verb` starts that length statistics miss. Intentional parallelism overrides it. |
-| Opening dominance | In a rolling 10-sentence window, flag any opening class occupying more than 60 percent | Governance default | Review trigger only. |
-| Clause-count variation | No more than 4 consecutive sentences with the same finite-clause count, unless the passage is deliberately parallel | Governance default | Catches syntactic monotony that survives length variation. |
+No instruction requires a long sentence, a short sentence after a long one, alternating lengths, a target distribution, a coefficient of variation, a tail value, or a repeated-opening quota. Sentence and paragraph shape follow the idea, audience, genre, and active voice.
 
-Where an authorized author profile exists, its measured corpus ranges replace every governance default above, because a measurement of that writer beats a default chosen for everyone. A writer whose corpus naturally runs wider than the spread row must not be normalized down into it, and a writer whose professional register is short briefing paragraphs must not be forced into long blocks.
+Review an ordinary sentence over 28 words. Keep it when splitting would damage a necessary relationship or exact meaning. Review repeated openings and identical paragraph shapes as possible patterns, not automatic failures. Buttons, headings, tables, lists, captions, code, identifiers, URLs, immutable quotations, legal text, and exact technical text keep their native form.
 
-**Boundary on the whole table:** these are controls for continuous prose. They do not apply to a button, a heading set, a table, a list, a caption, or a form label, and they are not evidence about how any reader responds.
+An authorized measured profile may describe the writer's habits. Use those measurements for optional comparison only. Do not generate toward a number or manufacture irregularity to resemble the corpus.
 
 ### The conflict this document does not resolve
 
 Two positions on cadence are in direct conflict, and neither side is strong enough to win.
 
-**The research position** is deliberate variance, expressed in the spread targets above.
+**The research position** is that measured variance can prompt editorial review.
 
 **The practitioner position** is uniform brevity. Experienced conversion copywriters advise short, choppy sentences, paragraphs of no more than two or three lines, and one-sentence paragraphs as the ideal, on scanned commercial surfaces.
 
 The most likely resolution is register: a landing page read by a scanning cold visitor is a different problem from expository prose read continuously, and the practitioners are talking about the first. That is a hypothesis. It has not been tested, neither position is graded above the level of an engineering heuristic, and this document adopts neither as universal.
 
-**What to do meanwhile:** apply the variance targets to continuous prose, apply the brevity preference to scanned commercial surfaces, and record which you applied if the choice is consequential. Do not report either as a finding, and do not resolve the conflict by quietly picking one.
+**What to do meanwhile:** let the idea and surface determine sentence shape. Use measured variance only as an optional diagnostic. Do not report either position as a finding.
 
 ## Open conflicts in this reference
 
@@ -562,4 +554,4 @@ Recorded so that neither side is lost. Do not close either by writing a rule.
 | Conflict | Position A | Position B | Status |
 |---|---|---|---|
 | Question-form subheadings | Practitioners reject them outright: give the answer instead of asking. | A direct observation in this project found question headings reading better than abstract noun-stack statement headings on the same page. | Open. The untested hypothesis is that questions beat noun stacks and lose to concrete answers. Neither side is graded above an assertion. |
-| Sentence cadence | Deliberate variance, per the targets above. | Uniform brevity on scanned commercial surfaces. | Open, probably register-dependent. Neither position is better than an engineering heuristic. |
+| Sentence cadence | Measured variance can prompt review. | Uniform brevity can fit scanned commercial surfaces. | Open, probably register-dependent. Neither position sets a generation quota. |

package/skills/agora/references/agora-marketing-runtime.md

@@ -0,0 +1,85 @@
+# Agora marketing runtime
+
+Use this compact file for ordinary `POSITION`, `SELL`, `INFORM`, and `TRANSACT` work. Use the deep marketing research reference only when the user asks for research-backed guidance, source review, an audit, or maintainer analysis.
+
+## Decide before planning
+
+Resolve these items in order:
+
+1. Required facts and exact wording.
+2. Deliverable, audience, genre, and length.
+3. Writing register.
+4. Persistent voice profile, task voice sketch, or plain default.
+5. Primary job and publication surface.
+6. The smallest useful argument.
+
+The default register is `PLAIN`. Choose `TECHNICAL`, `SCIENTIFIC`, `LEGAL_OR_COMPLIANCE`, or `AUDIT_OR_DILIGENCE` only when the requested job, audience, genre, or immutable text needs it. Topic words such as AI, software, data, security, engineering, and infrastructure do not choose the register.
+
+## Choose one primary job
+
+- `POSITION`: company profiles, About copy, category descriptions, and objective summaries.
+- `SELL`: marketing, sales, product pages, outreach, upgrades, and paywalls.
+- `INVEST`: fundraising, diligence, and capital decisions.
+- `INFORM`: editorial, educational, scientific, or technical explanation.
+- `TRANSACT`: buttons, forms, confirmations, alerts, and utility text.
+
+Choose the publication surface separately. Apply search and publication formatting only to indexable public work.
+
+## Preserve the brief
+
+Treat supplied product, offer, customer, result, price, term, route, permission, process, timing, legal, and outcome facts as the complete fact set unless the user authorizes invention.
+
+Do not invent names, figures, quotations, results, credentials, permissions, URLs, destinations, product behavior, or intermediate steps. Keep every material condition, limit, attribution, certainty level, and exact required term.
+
+Voice shapes sentence construction, paragraph shape, vocabulary, directness, openings, endings, rhythm, and where qualifications sit. Voice never changes facts or required wording.
+
+## Build the argument
+
+Start from the audience's live situation and next decision. Use only the moves the deliverable needs:
+
+```text
+situation -> stake -> useful difference -> how it works -> reason to believe -> next step
+```
+
+This is private planning, not a visible template. Very short copy may need only the live constraint, the strongest supplied difference, and the next action. Add detail only when it resolves a real doubt or enables action.
+
+For a very short company profile, a trigger or condition can lead when it makes the subject easier to understand. Retain the subject's primary supplied operation. Reordering may not reduce the company to an error state or downstream action.
+
+For very short sales copy, keep the exact constraint, conflict, threshold, or blocked action from the brief. Do not weaken a specific problem into generic complexity or challenges.
+
+Rank supplied facts by relevance to the reader's decision, differentiation, specificity, compression value, and omission risk. Preserve named limits, counts, coverage, entities, qualifications, and necessary lists. An inclusion statement does not communicate an excluded remainder when the exclusion is material.
+
+For customer-facing writing, name what the product or person does and what the reader can observe. Drop internal method names when the reader gains nothing from learning them.
+
+Preserve delivery ownership. Keep what the provider does, what the buyer supplies, the amount and timing of buyer work, dependencies, approvals, and coordination at the supplied level. Do not invent or magnify preparation, training, review cycles, handoffs, response times, or frequency.
+
+Do not add a plausible use case, reporting process, planning responsibility, reader behavior, comparison criterion, or benefit merely to make a section feel complete. Use only reader interpretations that follow directly from the supplied facts.
+
+When rewriting supplied copy, compare the candidate with the source for promise strength, certainty, buyer effort, provider responsibility, specificity, and next action. Keep the source when the candidate introduces a material regression or no material improvement.
+
+## Fit the channel
+
+- Heroes and ads: one recognizable stake, one meaningful difference, and the correct next action.
+- Product and comparison pages: explain enough for the reader to judge fit, terms, effort, and risk. Use prose unless the user asks for a specification table, field list, or checklist.
+- Company profiles: make the company easy to place and distinct without investor or audit language.
+- Professional email: put the request and needed context where they are easy to find.
+- Cold sales email: include a subject, greeting, brief body, and next step. When the copy is a reusable template and no recipient name is supplied, `[First name]` is an acceptable fill-in label.
+- Editorial work: give the reader useful reasoning before any commercial pressure.
+- Spoken work: use audible transitions, manageable clauses, and one clear ending.
+- Interface copy: name the state, action, object, or destination.
+
+For a button, prefer the immediate effect the user triggers. If selecting the button sends invitations, use `Send invitations` rather than a generic object label.
+
+If the brief supplies a required final shot, cancel consequence, negative capability, or exact product category, carry it into the requested component. Compression does not authorize omission.
+
+## Write calls to action
+
+Use a clear verb plus the concrete object, destination, or result. Match the commitment to what happens next. Do not invent a URL or destination.
+
+When no destination URL is supplied, return the action label as plain copy. Do not invent `#`, a dummy address, a route, or unresolved link markup.
+
+Keep one label for one materially identical action. If the brief names a report, form, preview, dashboard, demo, plan, or download, name it in the control or adjacent text.
+
+## Return the copy
+
+Return one recommended deliverable by default. Keep planning, fact checks, review notes, worksheet labels, and internal route names outside the ready-to-use copy unless the user requests them.

package/skills/agora/references/agora-marketing.md

@@ -819,73 +819,9 @@ Treat an existence-only route, screen, page, preview, or report as action availa
 
 ## Human voice and AI-writing-tell gate
 
-Apply this gate after the argument and proof pass. Its purpose is specific, accurate, author-owned prose. It does not prove human authorship or guarantee detector outcomes.
+This deep research reference does not restate the house list. The single canonical source is [human-voice-editing-reference.md](human-voice-editing-reference.md), and ordinary drafting follows [agora-writing-runtime.md](agora-writing-runtime.md).
 
-### Global output bans
-
-Hard invariant: emit zero U+2014 characters in the entire response. This applies to copy, headings, lists, critique, explanations, notes, metadata, quotations, and text copied from user or source material. Do not repeat U+2014 from an input. Replace it with a period, comma, colon, semicolon, parentheses, or plain hyphen as grammar requires.
-
-If an exact quotation contains U+2014, paraphrase it or state that it cannot be reproduced verbatim under the active constraint. Never alter a quotation and still present it as exact. After all drafting and editing, scan the complete response character by character for U+2014, replace every occurrence, and scan again. Return only when the count is zero.
-
-Unless immutable quoted, legal, technical, or user-required text must be preserved, also generate:
-
-- no curly or smart quotation marks;
-- no prompt acknowledgements or task meta-commentary;
-- no fabricated citations, quotations, anecdotes, memories, motives, credentials, or human texture;
-- no stock openings, canned contrast templates, or generic significance tails;
-- no unneeded conclusion or recap;
-- no deliberate errors, fake informality, or detector-evasion promises.
-
-Use straight quotes. Never change facts, names, figures, dates, citations, causality, uncertainty, legal meaning, or required technical terminology to satisfy a style preference. The U+2014 ban is not optional; resolve any conflict without misrepresenting altered source text as verbatim.
-
-### AI-heavy vocabulary
-
-Replace vague or inflated wording with concrete language when possible. Watch clusters of:
-
-- `delve`, `showcase`, `foster`, metaphorical `navigate`, `harness`, `leverage`, `unlock`, `unleash`, `empower`, `streamline`, `elevate`, `facilitate`, vague `optimize`;
-- `pivotal`, `crucial`, `robust`, `seamless`, `holistic`, `comprehensive`, `multifaceted`, `transformative`, `groundbreaking`, `cutting-edge`, `world-class`, `best-in-class`, `unparalleled`;
-- metaphorical `tapestry`, `landscape`, `realm`, `ecosystem`, `journey`, `testament`, `frontier`, `cornerstone`, `backbone`, `powerhouse`, `game-changer`;
-- `moreover`, `furthermore`, `additionally`, `notably`, `importantly`, `in essence`, `in summary`, `in conclusion`, `ultimately`, `that said`.
-
-These words are not scientific proof of machine authorship. Retain one when it is exact, technically required, part of a proper name, immutable source text, or more accurate than a forced synonym. Judge density, repetition, and vagueness rather than one ordinary word.
-
-### Stock templates and significance tails
-
-Delete or rewrite generated phrases such as:
-
-- `It is important to note that`;
-- `It is worth noting`;
-- `In today's fast-paced world`;
-- `In the ever-evolving landscape of`;
-- `At its core`;
-- `When it comes to`;
-- `A testament to`;
-- `Stands as a`;
-- `Plays a crucial role`;
-- `Not only X but also Y`;
-- `Not just X but Y`;
-- `Whether you are X or Y`;
-- `Unlocking the potential of`;
-- `Paving the way for`;
-- `Let me know if you would like`.
-
-Delete generic tails such as `underscoring the importance of`, `highlighting its role in`, `demonstrating its impact on`, `cementing its place as`, or `showcasing its commitment to`. End with information, action, or a supportable inference.
-
-### Structural tells
-
-Remove repeated `bold label: sentence` bullets, headings that restate themselves, identical paragraph shapes, conclusion sections that add nothing, balanced both-sides framing without named evidence, and smooth transitions that hide unrelated facts.
-
-Extend this to the whole deliverable, not one passage. Headings that share a syntactic template across a page, and calls to action that share a template across a site, are the same defect at a larger scale. A set of headings built as a count followed by a singular abstract noun, or every heading opening with the same conditional clause, is a structural tell even when each heading is accurate and each was written to a sanctioned shape. Sanctioned shapes are the most likely source of this failure, because nothing flags them individually.
-
-Avoid decorative tripartite rhetoric used as a stock flourish. Do not ban factual three-item series. Preserve necessary enumerations when they state scope, compare options, qualify a claim, meet a legal or platform requirement, or prevent misleading omission.
-
-Never add fake typos, fragments, slang, personal memories, jokes, or emotional confessions to simulate humanity. If author samples are supplied, follow stable habits without copying phrases or inventing an identity.
-
-### Detector limits
-
-Detector outputs are noisy, version-dependent, and vulnerable to domain shift. Never promise that writing is undetectable. Never treat a score as proof of authorship. If testing is authorized, compare against known-human samples from the same author and genre, use more than one method when feasible, record tool versions, and stop detector-driven editing when another pass no longer improves the prose independently.
-
-Research on detector limits includes the [RAID benchmark](https://arxiv.org/abs/2405.07940), [paraphrasing attacks](https://arxiv.org/abs/2303.13408), the [DAMAGE shared task](https://aclanthology.org/2025.genaidetect-1.9/), a [practical detector examination](https://aclanthology.org/2025.findings-naacl.271/), [bias against non-native English writers](https://arxiv.org/abs/2304.02819), [GPTZero methodology](https://arxiv.org/abs/2602.13042), [reliable detection limits](https://arxiv.org/abs/2303.11156), [human and AI boundary research](https://arxiv.org/abs/2502.15666), [flawed human heuristics](https://arxiv.org/abs/2206.07271), and [function-word adjacency networks](https://arxiv.org/abs/1406.4469).
+Resolve register and voice before argument planning. Apply the canonical hard rules after drafting without changing facts, names, figures, dates, citations, causality, uncertainty, legal meaning, or required technical terminology. Keep detector research and internal review language out of public copy. Never infer authorship or promise detector outcomes.
 
 ## Written GEO/AEO and citability
 
@@ -1468,7 +1404,7 @@ The full reference list, with lineage and adjudication for each entry, is mainta
 
 ### Human voice and detector evidence
 
-Use the detector sources in `Human voice and AI-writing-tell gate`. They support caution about detection claims, dataset shift, paraphrasing, and human bias. The absolute U+2014 character ban and the editorial bans on curly quotes, prompt leakage, canned templates, and generic significance tails are `HOUSE` rules. They are not proof of authorship.
+Use the detector sources in the canonical [human-voice editing reference](human-voice-editing-reference.md). They support caution about detection claims, dataset shift, paraphrasing, and human bias. The absolute U+2014 character ban remains an Agora invariant. The vocabulary, connective, prompt-leakage, template, significance-tail, typography, structure, and fidelity rules are output and editing rules, not proof of authorship.
 
 ### Preserved technical and research authorities
 

package/skills/agora/references/agora-science.md

@@ -22,15 +22,21 @@ User authority and the conflict hierarchy in [agora-marketing.md](agora-marketin
 
 ## Activate and route SCIENCE
 
-Activate `SCIENCE` when the asset explains, teaches, compares, summarizes, or persuades through scientific evidence or technical behavior. Select one internal route:
+Activate `SCIENCE` only when the requested deliverable needs research findings, study design, methods, statistics, uncertainty, scientific explanation, or evidence grading.
+
+Activate `TECHNICAL` only when the requested deliverable needs documented system behavior, engineering or systems explanation, architecture, API behavior, implementation detail, technical evaluation, or a technical audience.
+
+Do not activate either modifier because the subject or product category includes AI, software, data, security, engineering, infrastructure, or another technical label. A nontechnical homepage or product page in those categories stays in the `PLAIN` register.
+
+After the job and audience require a specialized explanation, select one internal route:
 
 | Route | Use for | Evidence focus |
 |---|---|---|
 | `EMPIRICAL` | Studies, experiments, observations, measurements, datasets, and research findings | Design, population, measurement, effect, uncertainty, and applicability |
-| `TECHNICAL` | Systems, interfaces, mechanisms, architecture, software, engineering, and failure behavior | Inputs, state changes, outputs, dependencies, constraints, and tradeoffs |
+| `TECHNICAL` | Systems, interfaces, architecture, API behavior, implementation, engineering, and failure behavior | Inputs, state changes, outputs, dependencies, constraints, and tradeoffs |
 | `MIXED` | An empirical result explained through a technical or scientific mechanism | Keep observed result and proposed explanation distinct |
 
-Choose the primary mode first. A public science explainer is normally `INFORM + SCIENCE`. A technical product page is `SELL + SCIENCE`. An investor memo about a research platform is `INVEST + SCIENCE`. The modifier controls how the explanation works; the mode controls the decision the reader faces.
+Choose the register and voice before the primary mode's argument plan. A public science explainer is normally `INFORM + SCIENCE`. An API guide is normally `INFORM + TECHNICAL`. A technical product section for engineers may use `SELL + TECHNICAL`. A general-audience homepage for the same product remains `SELL + PLAIN`. An investor memo using study results may use `INVEST + SCIENCE`. The modifier controls the requested explanation; the mode controls the decision the reader faces.
 
 Rule [HOUSE]: do not treat all technical writing as empirical science. An API contract needs exact documented behavior, not a research-study frame. Boundary: a technical claim supported by benchmark data uses `MIXED`, because its measured result still needs empirical qualification.
 
@@ -120,6 +126,14 @@ For a one-time observational measurement, state when temporal order is unknown a
 
 ### Preserve statistical meaning
 
+Describe a frequentist confidence interval as a range of values compatible with the data and analysis under the model assumptions. Do not translate it into the probability that the true effect lies in the interval or say the effect is likely above zero unless the stated statistical framework supports that probability claim.
+
+When the full interval lies above or below the null value, state that relation directly when it helps the reader interpret the result. Keep the estimated magnitude and uncertainty visible; do not stop at a significance label.
+
+Do not introduce `in practical terms`, call an effect modest or meaningful, or imply practical significance unless the brief supplies a practical threshold or outcome that supports that judgment.
+
+End on the supplied scientific limit. Do not append a generic recommendation to interpret cautiously, weigh benefits and harms, or use the result as a decision basis unless the task asks for that advice.
+
 Rule [A]: report magnitude and uncertainty, not a significance label alone. A small p-value does not establish effect size, importance, certainty, or a useful decision. See the [ASA statement on p-values](https://www.amstat.org/asa/files/pdfs/p-valuestatement.pdf) and [Cochrane's interpretation guidance](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-15).
 
 When material:
@@ -254,10 +268,14 @@ Use the strongest commercial claim requested by the user. Do not turn the hero i
 
 ### `CASE_STUDY + SCIENCE`
 
+Use scan-ready headings for implementation, validation result, limitation, role, and next decision when the supplied material supports them. If the material leaves both an external-validity gap and an unmeasured downstream outcome, keep them separate and make the next decision address each.
+
 Load [agora-case-studies.md](agora-case-studies.md). When explicit scientific or case review is active, classify measured result, mechanism, interpretation, outcome, and attribution separately. Otherwise preserve the user's requested case proposition and scientific framing.
 
 ### `VOICE + SCIENCE`
 
+When a next research step is requested, name the supplied or proposed design, comparator, measurement conditions, and uncertainty reporting. Do not turn a recommendation into `we will` without an approved plan.
+
 An authorized profile enters at hierarchy level 6. It changes expression, not proposition. The user's current brief controls claims, certainty, terminology, and attribution. Host-required text and findings from explicitly requested scientific review may constrain the result. Do not transfer a source author's opinions or certainty into a new claim.
 
 ### `GEO/AEO + SCIENCE`

package/skills/agora/references/agora-voice.md

@@ -1,12 +1,13 @@
 # Agora voice authority
 
-This reference governs `VOICE`: building a measured voice profile from a corpus, writing to that profile, and checking a draft against it. Load it only when the task builds, applies, inspects, or checks a voice profile.
+This reference governs two separate voice tiers: measured persistent profiles and task-only sketches from samples supplied with one task. Load it when the task builds, applies, inspects, or checks either tier.
 
-Voice is a modifier. It changes how a proposition is expressed. User authority in [agora-marketing.md](agora-marketing.md) controls profile use, and [agora-craft.md](agora-craft.md) supplies the rhythm controls a profile replaces. Agora does not authenticate corpus ownership, identity, attribution rights, endorsement rights, or publication permission. The user controls those decisions.
+Voice is a modifier. It changes how a proposition is expressed. User authority in [agora-marketing.md](agora-marketing.md) controls profile use. Numeric rhythm measurements remain optional diagnostics and never become drafting quotas. Agora does not authenticate corpus ownership, identity, attribution rights, endorsement rights, or publication permission. The user controls those decisions.
 
 ## Contents
 
 - [What VOICE is](#what-voice-is)
+- [Task-only voice sketches](#task-only-voice-sketches)
 - [Where profiles live](#where-profiles-live)
 - [Corpus admission](#corpus-admission)
 - [What gets measured](#what-gets-measured)
@@ -18,7 +19,7 @@ Voice is a modifier. It changes how a proposition is expressed. User authority i
 
 ## What VOICE is
 
-`VOICE` sits alongside `POSITION`, `SELL`, `INVEST`, `INFORM`, and `TRANSACT` with one difference: it is not mutually exclusive with them. Every other mode answers what job the copy is doing. `VOICE` answers whose habits the copy is written in, and it rides on top of whichever mode was already selected.
+`VOICE` sits alongside `POSITION`, `SELL`, `INVEST`, `INFORM`, and `TRANSACT` with one difference: it is not mutually exclusive with them. Every other mode answers what job the copy is doing. `VOICE` answers which observed habits shape the expression. Resolve register and voice before argument planning.
 
 The surface:
 
@@ -50,10 +51,33 @@ npx -p @maestroagora/agora agora-voice default --voice <slug>
 
 `voice build` is a measurement task, not a description task. A model asked to describe an author's voice writes flattery. The profile leads with numbers computed from the corpus, because a number is checkable and a later draft can be measured against it. Adjectives belong in the interpretation sections, underneath the measurements they interpret.
 
-**A profile the engine did not produce is not a profile.** Do not write one by reading a corpus and describing what you notice, and do not load a hand-written file as though it were measured. Where the engine cannot run, say so and work without a profile.
+**A persistent profile the engine did not produce is not a measured profile.** Do not load a hand-written file as though it were measured. A task-only sketch is permitted under the next section because it makes no certification, identity, or statistical claim and is never stored as a profile.
 
 **Boundary:** the measurement is only as good as the pipeline. Stylometric values move when the tokenizer, sentence segmenter, parser, or normalization rules change ([Grieve](https://doi.org/10.1093/llc/fqm020)). Freeze the pipeline in the profile and use the identical one when checking a draft. A comparison across two pipelines is not a comparison.
 
+## Task-only voice sketches
+
+A task-only sketch uses authentic samples supplied with the current task. It works without the persistent profile's 5,000-word certification floor because it does not claim stable identity or statistical matching.
+
+Use three to ten same-genre samples when available. Record a fuller local sketch only when those samples show recurring habits. With one or two samples, record cautious observations that are directly visible in the sample and mark internal confidence low. With no samples, preserve credible choices in the supplied draft and use plain professional writing.
+
+Record observable habits only:
+
+- sentence and paragraph range;
+- fragments and contractions;
+- first-person and second-person use;
+- punctuation;
+- directness, warmth, humor, and skepticism;
+- how qualifications are placed;
+- common opening, turning, and ending shapes;
+- constructions and words the samples avoid.
+
+The sketch has the internal type `TASK_VOICE_SKETCH`, a sample count, a genre label when supplied, and `certified: false`. It exists only for the current task. Never write it to `~/.agora/voices/`, add it to the profile index, or present it as a measured profile.
+
+Transfer habits, not material. Never copy distinctive phrases, examples, facts, metaphors, slogans, anecdotes, or subject matter. Never infer personality, biography, opinions, motives, identity, authorship, or approval. Run the existing exact overlap check and review unusual shorter matches before returning the draft.
+
+When task samples and a default persistent profile conflict, use the task samples for same-genre local expression in the current task. Keep only compatible stable habits from the default. An explicitly requested persistent profile remains active, but it still cannot authorize phrase copying or change the current brief.
+
 ## Where profiles live
 
 Profiles are stored at `~/.agora/voices/<slug>.md`, with `~/.agora/voices/index.json` as the registry. Never inside the skill directory.
@@ -157,7 +181,8 @@ concrete example.
 
 ## Vocabulary
 Owned: words and constructions recurring across genres, with frequency and one
-example line each. This list is an allowlist against the AI-vocabulary gate.
+example line each. This list documents author habit. It does not automatically
+override the priority anti-AI vocabulary gate.
 Avoided: words frequent in comparable writing and absent here. Record an
 avoidance only from a stated preference or a stable alternative in repeated
 eligible contexts. Absence alone is weak evidence.
@@ -187,7 +212,7 @@ Two sections carry more weight than they look.
 
 Keep supplied product actions at their stated scope. `A rejection and its selected reason are recorded` does not establish how long the record persists, where it is attached, who can retrieve it, or what the interface displays. Voice work may rephrase the supplied action. It may not fill operational gaps with plausible product behavior.
 
-**Transfer habits, not material.** Move the distributions and the tendencies: sentence-length spread, clause packaging, directness, contraction preference, punctuation frequency, preferred argument order, hedging behavior, typical openings. Do not move distinctive metaphors, slogans, signature phrases, anecdotes, factual examples, or source sentences. **Boundary:** the line is not always obvious. When a construction is both habitual and distinctive, abstract it before use: this author opens with a concrete contradiction is usable; reuse their particular contradiction is not.
+**Transfer habits, not material.** Use directness, contraction preference, punctuation habits, preferred argument order, qualification behavior, and typical openings before drafting. Keep measured sentence and paragraph distributions for comparison after drafting, not as targets. Do not move distinctive metaphors, slogans, signature phrases, anecdotes, factual examples, or source sentences. **Boundary:** the line is not always obvious. When a construction is both habitual and distinctive, abstract it before use: this author opens with a concrete contradiction is usable; reuse their particular contradiction is not.
 
 **Anti-mimicry checks.** After generating, flag every exact overlap of 8 or more consecutive tokens with the source corpus, excluding demonstrably generic strings, and review any shorter phrase that is unusual or repeatedly associated with the author. The token count is a **governance default** and an engineering review trigger; it is not a legal safe harbour, and no word count is one. Also review structural overlap: a sentence preserving the source's sequence of images and syntactic turns can be too close after every word has been swapped. **Boundary:** an overlap flag is a prompt to look, not a verdict. Common phrasing in a technical domain will trip it.
 
@@ -195,32 +220,33 @@ Keep supplied product actions at their stated scope. `A rejection and its select
 
 ## Voice against the tell gate
 
-Voice enters at **level 6** of the conflict hierarchy, with compression and channel fit. It never rises above it. Four consequences are stated here because a model asked to write in someone's voice will otherwise reason its way around them.
+Resolve voice before argument planning. Voice controls expression from the first outline, but never outranks required facts, exact wording, genre, audience, or first-read comprehension. Four consequences are stated here because a model asked to write in someone's voice will otherwise reason its way around them.
 
 1. **Voice follows the user's requested content.** Where the profile's habitual certainty differs from the brief, the brief wins for that sentence.
 2. **Voice never overrides host-required or user-required text.** Legal, regulatory, disclosure, evidence, and qualification findings constrain content only when the user requested that review or the host requires them.
 3. **Voice never overrides the U+2014 ban.** That constraint is immutable and sits at level 1. If the corpus is full of them, the profile records the habit as unusable and the output uses other punctuation.
-4. **Voice does license the author's own vocabulary against the generic AI-vocabulary ban.** This is a real exception and it needs writing down.
+4. **Voice identifies possible author-owned vocabulary but does not license it automatically against the priority anti-AI standard.**
 
-### The owned-vocabulary exception
+### The owned-vocabulary review
 
 The AI-tell gate and a voice profile genuinely conflict, and the conflict is not hypothetical. If an author writes `leverage` constantly across years and genres, the generic ban strips the word, the draft comes back in nobody's voice, and the feature the user paid for has been removed by a rule meant to make writing sound human.
 
-**The resolution: a profile's `## Vocabulary` owned list is an explicit allowlist that suppresses the generic AI-vocabulary ban for those specific words, and only those.** Everything not on the list stays banned. The allowlist is scoped to the profile, not to the session, and it does not travel to work written without that profile.
+**The resolution: a profile's `## Vocabulary` owned list is evidence for a narrow review, not an automatic allowlist.** Retain a listed banned word only when it is also load-bearing, exact, technically required, part of a verified proper name or immutable text, or explicitly required by the current user or house style. Measurement alone is insufficient.
 
-Three conditions bound it:
+Four conditions bound the review:
 
 - **The word has to be on the list because it was measured.** A word recorded from frequency across genres qualifies. A word added because the draft wanted it does not.
-- **It suppresses the vocabulary ban only.** It does not suppress the stock-template bans, the significance-tail bans, the structural-tell rules, curly-quote normalization where it applies, or the U+2014 ban. Those are not vocabulary.
-- **It does not change the proposition by itself.** `Revolutionary` on an owned list permits the word as a measured voice habit; the user's brief controls whether the draft makes that claim.
+- **The word must carry exact meaning.** Familiarity or voice match alone does not make a banned term load-bearing.
+- **Any exception reaches vocabulary only.** It does not suppress the stock-template bans, significance-tail bans, structural-tell rules, curly-quote normalization, or the U+2014 ban.
+- **It does not change the proposition by itself.** `Revolutionary` on an owned list records a measured habit; the user's brief controls whether the draft makes that claim, and the priority anti-AI standard controls whether the wording survives.
 
-Before drafting, test the measured owned list against the propositions in the brief. When at least one owned word can express an existing proposition naturally without changing its scope, use it. Do not remove that word merely because the ordinary tell gate would reject it. Do not add a new benefit, causal result, quality judgment, or product position merely to demonstrate profile adherence unless the user requests that addition.
+Before drafting, test the measured owned list against the propositions and the priority anti-AI standard. Keep a banned owned word only when the narrow exception is documented. Do not add a new benefit, causal result, quality judgment, or product position merely to demonstrate profile adherence.
 
-For a production profile whose brief explicitly supplies measured owned vocabulary, treat that test as a required pre-draft step. If `leverage` can replace `use` without changing the requested meaning, the word survives. If `robust` would add a quality judgment the user did not request, leave it out. Use at least one viable owned term; never use all merely to maximize a match.
+For a production profile whose brief explicitly supplies measured owned vocabulary, treat that review as a required pre-draft step. If `leverage` merely replaces `use`, remove it. If a required technical term, verified name, immutable quotation, or explicit user instruction requires the word, preserve it and record why. Never use a banned owned term merely to maximize a match.
 
-Sentence-length and paragraph-shape measurements are distributions, not quotas or stock structures. Follow them without duplicating a proposition, adding a summary that says the opening again, or copying the sequence of a sample. The structural-tell gate fails identifiable canned framing, significance tails, decorative triads, and repeated stock templates. It does not fail a draft merely because its sentence and paragraph lengths resemble the authorized profile.
+Sentence-length and paragraph-shape measurements are diagnostics, not quotas or stock structures. Do not generate toward them, add a long sentence to satisfy a tail, alternate lengths, force paragraph variance, duplicate a proposition, add a summary, or copy the sequence of a sample. The structural-tell gate fails identifiable canned framing, significance tails, decorative triads, and repeated stock templates. It does not fail a draft merely because its sentence and paragraph lengths resemble the authorized profile.
 
-**Boundary:** an owned-vocabulary list assembled from a thin corpus is a licence built on noise. Below the certification floor, no allowlist is issued at all, because a word cannot be shown to recur across genres in a corpus that has one.
+**Boundary:** an owned-vocabulary list assembled from a thin corpus is a preference claim built on noise. Below the certification floor, no candidate exception list is issued, because a word cannot be shown to recur across genres in a corpus that has one.
 
 ## Checking adherence