Brand Agency

Brand Agency

Run a large-scale brand engagement through ten research workstreams, iterative cross-team debate, and seven implementation-ready deliverables. Ground decisions in client evidence, current research, and explicit trade-offs.

Non-negotiable principles

• Treat every project as unique. Never reuse a sector assumption without evidence.
• Match the user's language in client-facing interaction and project outputs. Keep the skill source itself in English.
• Use the current execution date. Never rely on hard-coded trend periods.
• Respect the host agent's policies, permissions, tool names, and safety boundaries.
• Never request or expose private chain-of-thought. Record concise rationales, evidence, uncertainties, and decisions instead.
• Apply the built-in Beast Mode Ultra protocol to every substantial research, debate, synthesis, production, and audit role.
• Write each phase to disk so the engagement remains auditable across long or multi-turn runs.
• Keep generated project artifacts in a dedicated branding output directory at the active project workspace root, never inside skills/brand-agency/ or any installed skill directory.
• Keep the minimum length of every research report at 4,000 words and every final deliverable at 7,000 words unless the user explicitly requests a shorter edition.

Reference routing

Read only the files required for the active phase:

Phase 0: Assess capabilities and scope

Read references/agent-operating-guidelines.md and references/beast-mode-ultra.md.

1. Identify available capabilities: filesystem access, interactive questions, web research, document inspection, image inspection, specialist workers or subagents, and parallel execution.
2. Activate Beast Mode Ultra as an inline operating protocol. Do not require it to be installed as a custom agent.
3. Use capability-based behavior rather than requiring a named tool or agent.
4. If specialist workers are available, assign independent workstreams and run safe tasks in parallel within host limits. Ensure each worker reads or receives the complete Beast Mode Ultra protocol.
5. If specialist workers are unavailable, execute the same workstreams sequentially as clearly separated Beast Mode Ultra roles and save each report before starting the next.
6. If web research is unavailable, proceed from supplied evidence, label external-research gaps, and never fabricate sources.
7. Confirm that the request is a large engagement. If it is materially smaller, explain the scale of this workflow and offer a scoped alternative.

Phase 1: Build the project context

Read references/client-interview.md.

1. Inspect every brief, asset, prior guideline, research file, or constraint supplied by the user.
2. Ask the ten strategic questions. Use an interactive question tool if available; otherwise ask in chat in manageable batches.
3. Do not ask for information that the user already provided.
4. Select the canonical output root:
   • prefer branding-<project-slug>/ when the project or entity has a stable name;
   • use branding/ only when no usable project name exists or the user explicitly requests it;
   • derive <project-slug> from user-provided project or entity wording, normalize it to lowercase ASCII kebab-case, and do not invent a name;
   • place the directory at the active project workspace root, outside the installed skill directory;
   • never merge with a directory belonging to another engagement; use a user-approved distinct slug when a collision is ambiguous.
5. In every path in this skill and its references, replace <branding-root> with that selected directory name. Never create a directory literally named <branding-root>.
6. Create <branding-root>/context/client-brief.md.
7. Create <branding-root>/context/engagement-plan.md, including:
   • canonical output root and project slug;
   • entity type and maturity;
   • engagement objective;
   • audiences and stakeholders;
   • markets, cultures, and languages;
   • relevant touchpoints and operating environments;
   • existing equity and constraints;
   • actual budget, timeline, and resources when known;
   • decision makers and approval process;
   • evidence gaps and assumptions;
   • selected adaptations for each workstream and deliverable.

Do not invent a budget, timeline, business model, distribution channel, technology stack, or launch stage.

Phase 2: Create the canonical workspace

Use these project-relative POSIX-style paths on every operating system:

<branding-root>/                  # branding-<project-slug>/ preferred; branding/ fallback
|-- context/
|   |-- client-brief.md
|   `-- engagement-plan.md
|-- research/
|   |-- team-01-context-ecosystem.md
|   |-- team-02-audiences-stakeholders.md
|   |-- team-03-organization-strategy.md
|   |-- team-04-category-culture-semiotics.md
|   |-- team-05-verbal-identity.md
|   |-- team-06-visual-identity.md
|   |-- team-07-experience-touchpoints.md
|   |-- team-08-communications-channels.md
|   |-- team-09-activation-change.md
|   `-- team-10-governance-risk-measurement.md
|-- debate/
|   |-- round-01-conflict-map.md
|   |-- round-02-resolution.md
|   |-- round-03-red-team.md
|   |-- round-N-targeted-resolution.md
|   `-- decisions.md
`-- deliverables/
    |-- 00-deliverables-index.md
    |-- 01-brand-strategy-positioning.md
    |-- 02-audience-stakeholder-system.md
    |-- 03-verbal-identity-messaging.md
    |-- 04-visual-identity-logo-system.md
    |-- 05-brand-experience-applications.md
    |-- 06-communications-activation.md
    `-- 07-implementation-governance-measurement.md

Create only these paths under the selected output root. Keep skills/brand-agency/ read-only: it contains reusable instructions and resources, not engagement output.

Phase 3: Run ten research workstreams

Read references/team-prompts.md.

Run all ten workstreams for a full engagement under Beast Mode Ultra. Adapt their research questions to the entity:

Batch by dependency:

• Batch A: Teams 01-04.
• Batch B: Teams 05-07 after Batch A.
• Batch C: Teams 08-10 after Batch B.

Each report must:

• contain at least 4,000 words;
• use the client context and preceding reports where relevant;
• distinguish facts, interpretations, recommendations, and unknowns;
• cite sources with publication date and access date when web research is available;
• avoid invented quotations, statistics, or research participants;
• end with ten strategic insights and implications for the debate.

Phase 4: Run iterative debate

Read references/debate-protocol.md.

Run every debate and red-team role under Beast Mode Ultra. Run at least three rounds:

1. <branding-root>/debate/round-01-conflict-map.md: identify agreements, conflicts, unsupported claims, and missing evidence.
2. <branding-root>/debate/round-02-resolution.md: conduct cross-workstream critique, compare options, and make provisional decisions.
3. <branding-root>/debate/round-03-red-team.md: challenge the provisional system for coherence, distinctiveness, feasibility, accessibility, cultural fit, risk, and long-term resilience.

Run additional targeted rounds whenever a critical decision remains unresolved, new evidence changes a decision, or the red team finds a material defect. Finish only when the convergence criteria in the protocol are met or a clearly documented client decision is required.

Save the binding decision register to <branding-root>/debate/decisions.md. Treat recorded votes as a synthesis method, not as empirical validation.

Phase 5: Produce seven final deliverables

Read references/document-templates.md.

Create all seven deliverables under Beast Mode Ultra for a full engagement and tailor every application section to the actual entity and environments. Do not insert app screens, code, social channels, packaging, signage, investor materials, or launch mechanics unless they are relevant.

Each final deliverable must:

• contain at least 7,000 words;
• cite the research and decision-register sections that support major choices;
• include implementation-ready specifications appropriate to its medium;
• identify owners, dependencies, risks, and open decisions;
• remain consistent with all other deliverables;
• contain no unfilled placeholders.

Create <branding-root>/deliverables/00-deliverables-index.md after the seven documents. The index has no minimum length.

The final implementation and stewardship deliverable must be saved to <branding-root>/deliverables/07-implementation-governance-measurement.md.

Phase 6: Final audit

Apply the final Beast Mode Ultra validation matrix.

Verify:

• all required files exist at the canonical paths;
• every report and deliverable meets its minimum length;
• every major decision is traceable to evidence and <branding-root>/debate/decisions.md;
• names, colors, terminology, architecture, and priorities are consistent;
• culturally or legally sensitive claims are qualified and assigned for expert review;
• no unsupported project assumptions survived from templates;
• no tool-specific instruction leaked into client deliverables;
• no mojibake, replacement characters, hidden credentials, or personal filesystem paths remain.

Report completion, unresolved client decisions, evidence limitations, and recommended next actions.