Maestro: Agora

Say what matters. Make it land.

Agora helps you turn what your business does, knows, sells, imagines, or explains into writing people can understand and act on.

It starts with the decision behind the copy. Who must understand, believe, choose, approve, fund, or do something next? Agora finds the consequential stake, explains why the subject matters, chooses the details that move the decision, and writes for the actual channel.

Use it for landing pages, heroes, ads, product copy, sales outreach, investor communication, scientific and technical explanation, case studies, company profiles, editorial work, interface text, and spoken scripts. Agora follows your brief. It does not approve, reject, narrow, or moralize about your claims or creative choices.

What Agora adds

Agora returns one ready-to-use result by default. Alternatives, internal routes, and rationale stay out of the final copy unless requested.

When buyers ask AI, is your brand part of the answer?

More decisions now start inside an AI answer. Buyers ask for comparisons, explanations, and recommendations; brands either enter that answer or they do not.

CiteSurge tracks how your brand and competitors appear across ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews, Bing Copilot, and Grok. Run Prompt Scans for the questions your customers may ask. Measure competitive Share of Voice. See which brands and sources shape each answer. Then turn what you find into prioritized Content Recommendations and keep monitoring the same prompts as the competitive picture changes.

Find out where your brand stands when buyers ask AI.

Install

Install Agora across the shared Agent Skills path and Claude Code:

npx -y @maestroagora/agora

The default user install writes the same reviewed skill to:

• ~/.agents/skills/agora for Codex and Agent Skills-compatible tools.
• ~/.claude/skills/agora for Claude Code.

Open a new task or restart the host after installation so its skill registry and slash menu reload.

Install one target or use project-local scope:

npx -y @maestroagora/agora --target codex --scope user
npx -y @maestroagora/agora --target cursor --scope project
npx -y @maestroagora/agora --target codex,claude --scope user
npx -y @maestroagora/agora --target universal --dry-run

Supported targets are universal, shared, codex, claude, cursor, gemini, copilot, and windsurf. Add --force only when you intend to replace a different copy at the exact agora destination.

Update an existing user installation:

npx -y @maestroagora/agora@latest --target universal --scope user --force

Native plugin install

Claude Code:

/plugin marketplace add mbanderas/maestro-agora
/plugin install maestro-agora@maestro-agora

Codex CLI:

codex plugin marketplace add mbanderas/maestro-agora
codex plugin add maestro-agora@maestro-agora

The npm installer provides the broadest host coverage. Native plugin commands use the matching Claude or Codex manifest from this repository.

Quick start

Invoke Agora directly and provide the brief, material, and claims you want it to use:

/agora Rewrite this upgrade screen. Make the blocked action matter, state the plan difference clearly, and use one supported CTA.

Choose a primary mode when you want to override inference:

/agora position Turn these business facts into a 35-word company profile.
/agora sell Build a homepage hero around the strongest buyer stake in this brief.
/agora invest Write a fundraising opening from the current results, risks, and purpose of the round.
/agora inform Explain this research finding for a public audience.
/agora transact Rewrite this confirmation so the state and next action are unmistakable.

Add modifiers when the subject or asset needs them:

/agora sell technical Write a product section for engineers with this documented behavior.
/agora inform science Explain this study for a general audience.
/agora sell case study Write a customer success case from this approved project record.
/agora inform science case study Explain this engineering implementation and its measured limits.
/agora invest science Build a deep-tech capital case from the supplied study and scale-up results.
/agora invest case study Use this customer case to strengthen the investor brief.
/agora sell --voice house Rewrite this page in the measured house profile.

In Codex, $agora and the skills picker can also select the installed skill. Other hosts may expose skills through a picker or mention syntax. Asking the agent to use the Agora skill remains portable.

Article generation

For articles, Agora starts with your instructions, audience, voice, and supplied material. It uses first-party observations, customer examples, numbers, decisions, and constraints to plan the argument before drafting. Sections receive space according to what the material can support. A requested outline or style still controls the result. Sparse briefs produce useful explanations without invented customer stories, personal experiences, opinions, or statistics. The final pass removes generic repetition and internal words such as evidence, claim, and qualification when ordinary readers do not need them. Agora states what happened or who said it in everyday language and preserves your facts and intended message. This happens in the existing writing flow and returns a ready-to-use draft.

When Node is available, Agora can use its shipped article typography check during the final writing pass. It reports character positions for the writer to correct and does not score, classify, or rewrite the article.

Human-voice editing standard

Agora ships the attached source unchanged as human-voice-editing-reference.md. It is the single canonical authority for banned vocabulary, connectives, templates, significance tails, punctuation, prompt leakage, structural tells, author samples, meaning preservation, genre fit, detector limits, and privacy cautions.

The standard removes generated stock vocabulary, connective phrases, templates, significance tails, prompt leakage, repeated structural patterns, raw channel residue, fake human texture, and detector-driven corruption. It preserves the current user's facts, required wording, technical terms, genre, and intended meaning. Measured voice profiles document author habits but do not automatically exempt banned vocabulary.

Plain professional writing is the default. AI, software, data, security, engineering, and infrastructure products stay in plain language for ordinary readers. A specialized register activates only when the deliverable needs scientific findings, technical detail, legal wording, or formal review.

Run the deterministic style review on a file:

npx -y -p @maestroagora/agora@latest agora-style-audit ./draft.md
npx -y -p @maestroagora/agora@latest agora-style-audit ./draft.md --register technical --json

Hard findings cover banned typography, canonical vocabulary and templates, prompt leakage, and significance tails. Review warnings cover plain-register control words, long or joined sentences, repeated openings and paragraph shapes, noun and preposition stacks, false reframes, question fragments, corporate helper phrases, and legalistic leakage. The tool does not rewrite text, validate facts, infer authorship, or promise detector evasion.

Publication privacy and provenance

Agora writes copy. It does not create SVG, PNG, JPEG, PDF, DOCX, or PPTX files by itself. Other document, presentation, PDF, site, design, or image tools may place Agora copy into those artifacts.

When you explicitly request a publication audit, Agora can inspect the local artifact before it leaves your control:

/agora publication audit Inspect ./launch-deck.pptx before I share it.

Run the packaged read-only inspector directly when you want a deterministic report:

npx -y -p @maestroagora/agora@latest agora-publication-audit ./launch-deck.pptx
npx -y -p @maestroagora/agora@latest agora-publication-audit ./public-assets --json --output audit.json

The inspector covers configured hidden or control characters in text, selected HTML and SVG metadata, Office properties and review material, PDF metadata, PNG and JPEG metadata containers, and C2PA carrier hints. ExifTool adds broader read-only PDF and image metadata coverage when it is already installed. c2patool verification is opt-in through --verify-c2pa, with remote-manifest fetching disabled by supplied settings.

Source files are read only and compared by SHA-256 before and after inspection. Sensitive values and absolute paths are redacted by default. Missing or partial coverage is reported as UNKNOWN.

This is publication hygiene, not a watermark remover or authorship detector. A result cannot prove that content is AI-generated, human-written, anonymous, clean, or safe. Claude's documented model-level text marking is not equivalent to unusual Unicode or ordinary file metadata.

Model-level text watermarks are statistical generation signals, not hidden characters or metadata fields. Without the matching provider verifier or detector configuration and enough eligible text, Agora reports that question as uninspected instead of guessing from style, Unicode, or generic detector scores.

Routing model

Agora resolves the audience, genre, writing register, and voice before it plans the argument. It then chooses a primary mode, publication surface, and any domain or asset modifiers. Voice shapes expression from the first sentence while preserving required facts and wording.

Primary modes

POSITION remains the default for descriptive company profiles, even when investors may read them. Directory placement does not convert an objective profile into a capital pitch.

Composable modifiers

The modifiers can compose. A technical fundraising case may use INVEST + TECHNICAL + CASE_STUDY. A founder-voiced scientific product video may use SELL + SCIENCE + VOICE + HYBRID.

Subject matter alone does not activate SCIENCE or TECHNICAL. A nontechnical homepage for an AI or security product remains SELL + PLAIN.

Surfaces

Titles, descriptions, captions, show notes, transcripts, and companion pages receive written treatment. Spoken narration stays free of search-format scaffolding.

How the argument works

Agora reasons through a variable-depth path:

situation -> stake -> criterion when useful -> mechanism -> reason to believe -> destination belief -> next step

This path is not a paragraph template.

• Short copy pairs the strongest live consequence with the strongest supplied differentiator.
• Medium copy adds the mechanism and the result, detail, or qualification that matters most.
• Long copy expands only when another fact resolves a real objection or expensive uncertainty.

Agora ranks supporting details by decision relevance, differentiation, specificity, compression value, and omission risk. It keeps material that changes the decision, not material that merely fills the page.

Heroes and short-form sales copy

A hero is one distributed argument, not a headline contest. Agora evaluates:

• Optional eyebrow.
• Headline.
• Subhead.
• Primary and optional secondary CTA.
• Result or qualification microcopy.
• Visual, demo, or result context.
• Immediate next section.

It distinguishes brief fidelity from persuasive quality. Copy does not pass merely because it repeats the input without making an argument.

SELL on a first-attention surface normally receives a commercially assertive treatment. Mid-funnel explanation receives a persuasive-explanatory treatment. Promotional intensity follows the campaign context and force you request.

Promise grammar matters. See X, Learn how to X, We help you X, Do X more often, You will X, and an imperative such as Beat X create different expectations. CTA language must match the destination and its commitment level.

Scientific and technical communication

Specialized explanation supports three internal routes:

When you request scientific claim review, Agora can classify material as observation, established fact or consensus, model, interpretation, implication, recommendation, hypothesis, speculation, or unknown. Without that request, it follows the certainty and framing in your brief.

Agora can preserve study design, population, comparator, baseline, denominator, absolute and relative effects, material uncertainty, and the difference between statistical and practical significance when your task needs that precision.

Misconception hooks, question-first explanation, analogies, visuals, and alternating narrative threads are optional tools. Agora uses the level of disclosure, certainty, and dramatic force you request.

Case studies built around what happened

CASE_STUDY supports:

Case family and project status are separate:

You choose whether a case is real, fictional, mixed, anonymous, attributed, disclosed, or presented without a status label. Agora does not police that choice.

Every result is classified before writing: measured outcome, customer-reported outcome, observed process or adoption change, supported inference, target, pending measurement, or unmeasured.

Agora can separate delivery activity, business impact, chronology, and causality when you request case review. Otherwise it follows the outcome framing in your brief.

Permission, attribution, confidentiality, quote, and disclosure review are optional. Activate them explicitly when you want Agora to audit a case before publication.

Academic and clinical case reports remain outside this capability.

Investment communication

INVEST remains one primary mode with three internal routes:

Agora can separate historical results, current state, contracted commitments, customer reports, forecasts, targets, model assumptions, interpretations, scenarios, and unknowns when you request diligence or claim review. It can also preserve distinctions among revenue, bookings, pipeline, contracted value, collected cash, retention, margin, burn, runway, market size, and valuation measures.

It can produce warm introductions, direct outreach, spoken openings, decks, briefs, demos, follow-ups, diligence memos, investment-committee briefs, data-room narratives, and progress updates. It does not impose a universal deck order, deck length, meeting script, talk ratio, outreach rule, or opening duration.

Strong fundraising communication makes the decision, present position, purpose of the capital, and next milestone clear. Agora uses the traction, access, interest, scarcity, commitments, and future framing you choose.

When diligence is the task, Agora can keep a weakness visible until new results resolve it. In fundraising work, it follows the strategic framing you request.

Legal, securities, offering, solicitation, eligibility, and disclosure review is optional and must be requested explicitly.

Measured voice profiles

VOICE modifies another mode rather than replacing it. Persistent profiles are measured from the corpus you select, not improvised from adjectives.

Build and inspect a profile with the shipped engine:

npx -p @maestroagora/agora agora-voice build --name house \
  --register blog --from ./posts \
  --register email --from ./letters

npx -p @maestroagora/agora agora-voice list
npx -p @maestroagora/agora agora-voice check --voice house ./draft.md

The engine reads Markdown, plain text, and HTML from files, directories, and URLs. It removes quotations, code, tables, signatures, front matter, and headings before measurement. Binary document formats are refused by name rather than partially extracted.

A profile records sentence and paragraph distributions, function words, punctuation, openings, stance, contraction behavior, and supported vocabulary. It states what the corpus was too small to measure and refuses certification below 5,000 clean author-controlled words.

Profiles live at ~/.agora/voices/, outside the replaceable skill directory. A default profile can apply across modes. Use --no-voice, neutral, or --voice <name> per request. Your required wording overrides habitual profile tendencies.

Task-only voice sketches are separate. When you supply authentic samples with one task, Agora can use three to ten same-genre samples without the 5,000-word persistent-profile floor. One or two samples support only cautious local observations. The sketch records visible habits, stays inside the task, is never certified or stored as a profile, and never claims authorship or statistical identity. Agora does not copy distinctive phrases, facts, examples, metaphors, slogans, or anecdotes from the samples. With no samples, it uses plain professional writing and preserves credible choices from your draft.

Sentence and paragraph measurements remain available in voice check. They are diagnostics, not generation quotas. Agora does not force a 25-word sentence, alternate sentence lengths, or write toward a target distribution.

Agora builds or applies the profile you request. You are responsible for corpus rights, identity use, attribution, endorsements, disclosure, and publication.

User control and responsibility

Agora is a writing tool, not a content moderator, fact checker, approval system, or legal reviewer. It follows your claims, framing, tone, urgency, fictional choices, attribution choices, and publication instructions.

• Agora does not refuse, narrow, weaken, qualify, remove, or label content because it considers a claim unsupported or a writing choice too aggressive.
• Agora does not append warnings, moral judgments, policy notes, or generic disclaimers to requested copy.
• Fact checking, source review, evidence review, permission review, confidentiality review, disclosure review, due diligence, and compliance review run only when you request them.
• Host platforms and model providers may still apply their own terms and safety policies.
• You are responsible for reviewing and using every output, including its accuracy, legality, claims, permissions, confidentiality, attribution, disclosures, and compliance with platform rules.

Read User Responsibility and Disclaimer and Privacy Notice.

Written GEO/AEO boundaries

For indexable public assets, Agora answers the reader's question early when the format calls for it, names entities and scope, keeps supporting details and qualifications beside claims, exposes supplied provenance, and builds useful passages that remain accurate when quoted alone.

For indexable public pages, it can flag relevant crawlability, canonical, sitemap, metadata, structured-data, accessibility, and delivery checks. These practices may improve eligibility and citability. They cannot promise retrieval, selection, quotation, citation, ranking, recommendation, referral, conversion, or revenue.

GEO/AEO applies to coherent passages on indexable public pages, not every sentence. It does not automatically apply to private, non-indexable, or spoken work, and it cannot force article-style density into a hero, case-study opening, or short CTA.

Package architecture

The shipped skill stays progressively loaded:

skills/agora/
|-- SKILL.md
|-- agents/
|   `-- openai.yaml
|-- scripts/
|   `-- publication-audit.mjs
`-- references/
    |-- agora-writing-runtime.md
    |-- agora-marketing-runtime.md
    |-- agora-conversion-runtime.md
    |-- agora-case-study-runtime.md
    |-- human-voice-editing-reference.md
    |-- agora-case-studies.md
    |-- agora-conversion.md
    |-- agora-craft.md
    |-- agora-invest.md
    |-- agora-marketing.md
    |-- agora-publication.md
    |-- agora-science.md
    `-- agora-voice.md

SKILL.md contains routing. Ordinary work loads the compact writing and job runtimes first. Deep research stays available for explicit research, source review, scientific, technical, legal, audit, diligence, and maintainer work.

• human-voice-editing-reference.md is the exact canonical human-voice editing authority.
• agora-writing-runtime.md is the small plain-language writing contract loaded for every writing task.
• agora-marketing-runtime.md, agora-conversion-runtime.md, and agora-case-study-runtime.md keep ordinary drafting out of the research register.
• agora-marketing.md preserves deep doctrine, source material, research grades, and maintainer guidance.
• agora-craft.md adds headlines, heroes, awareness and sophistication, emotion, and prose rhythm.
• agora-science.md adds empirical and technical explanation plus optional scientific claim review.
• agora-case-studies.md adds case structure, results, and optional attribution, permission, and confidentiality review.
• agora-conversion.md adds bounded conversion priors, outcome matching, self-serve and enterprise route design, pricing decision contracts, decision-adjacent proof, contradiction handling, and experiment interpretation for funnel surfaces.
• agora-invest.md adds fundraising, diligence, allocation, and optional claim-review procedures.
• agora-publication.md adds explicit read-only publication privacy and provenance review for local artifacts.
• agora-voice.md adds measured voice profiles and user-controlled profile use.
• publication-audit.mjs produces a deterministic, source-preserving inspection report without a cleaning operation.

Public-package hygiene

Research informed Agora, but research custody is separate from distribution.

The public repository and npm package must not contain raw or corrected transcripts, caption files, supplied PDFs or office documents, audio, video, private source identities, model-output scratch, research working files, local paths, or assigned secrets.

npm run check runs validation, deterministic tests, the exact package allowlist, and release hygiene. npm run release:check adds the current frozen blind-evaluation evidence gate. npm pack and npm publish invoke it through prepack and prepublishOnly.

evals/releases/current.json is the explicit release contract. Its version must match package.json, the frozen manifest, schemas, release gates, adjudications, records, custody hashes, and external artifact manifests. Missing or stale human-writing evidence blocks a publishable release. Live generation does not run inside unit tests.

The versioned directories under evals/blind/ are public pairwise-release artifacts, not permanently secret holdouts. v1.2.0, v1.4.0, v1.5.0, and v1.7.0 are frozen by exact tree hashes in evals/releases/locks.json; validation fails on additions, deletions, or edits.

evals/prospective/article-generation-v1.0.0/ contains the article-generation development cases for this release. The frozen current manifest lives under evals/blind/ after prospective generation and blind pairwise review. Earlier fixture sets remain frozen for regression analysis and historical reproducibility.

Deterministic tests verify instruction structure, routing contracts, static invariants, and evaluation-record shape. Versioned blind-evaluation tooling remains available for repeatable development analysis without turning model preference into a universal conversion claim.

Verify

npm run check
npm run release:check
node scripts/eval-locks.mjs
npm pack --dry-run --json
npx -y @maestroagora/agora --dry-run

The mandatory release gate checks skill structure, routing contracts, factual and user-authority boundaries, typography, metadata, reference links, installer parity, exact npm contents, frozen evaluation-tree locks, public-tree hygiene, deterministic tests, and checked current blind evidence. Generation and judging happen before evidence is frozen, not during routine tests.

Change record

Every research-derived rule records its source grade and boundary or failure condition. Numeric thresholds are sourced or labeled as governance defaults. Practitioner procedures remain bounded craft guidance, not scientific or commercial performance laws.

Maestro suite

• Maestro Frontier: Fuses the model CLIs you already run into one judged, grounded answer.
• Maestro Agora: Writes persuasive copy, technical explanations, compelling case studies, and investment communication to your brief.
• Maestro CostGuard: Audits CI and cloud infrastructure for cost leaks and shows what to fix.

License

MIT. See User Responsibility and Disclaimer and Privacy Notice.

{
  "name": "maestro-agora",
  "description": "Self-hosted marketplace for Maestro: Agora, the argument-first writing and publication artifact review skill.",
  "owner": {
    "name": "Mark Laursen",
    "url": "https://github.com/mbanderas"
  },
  "plugins": [
    {
      "name": "maestro-agora",
      "source": "./",
      "description": "Write user-directed commercial, technical, case-study, investment, interface, and spoken arguments, then inspect local publication artifacts when explicitly requested."
    }
  ]
}