🧹 Solution Debranding

Three agent skills that coordinate a plan → apply → verify workflow for removing organization-specific identity and coupling from a codebase, so the result is portable enough to reuse, hand over, or release publicly. You get a scan of everything that needs attention, a plan that tracks approvals and evidence, changes applied one unit at a time, each validated before the next, and an independent audit of the result.

Renaming identifiers only addresses visible identity. It does not remove the assumptions and dependencies that bind a solution to its original organization or environment. Plan mode's coupling review finds them and routes each one to the track that owns it.

When it fits · How it works · Install · Try it · Safety · FAQ · Contributing

When it fits

Portability and reuse

• Reuse across brands. Ship without the first brand's name, tenant IDs, or endpoints, and find the infrastructure and integrations that still assume the original owner
• White-label a product. Load each deployment's logo, palette, and name from configuration instead of hardcoded values
• Extract a shared platform. Find where the original tenant leaked into common code

Handover and ownership transfer

• Hand over or open-source. Remove owner-specific tooling, planning records, credentials, and environment assumptions before transfer
• Offboard a contractor. Remove references, agent configuration, and planning artifacts left in the codebase
• Transfer between teams. Remove hostnames, distribution lists, and on-call details owned by the original business unit

Rebranding and design

• Adopt a design system. Rebrand through Fluent 2, Carbon, Material, or another system using its supported theming contract

Compliance and release

• Assess public-release readiness. Check Git history, external surfaces, datasets, and documentation
• Prepare ownership-change evidence. Record embedded third-party brands, credentials, tenant identity, and owner-specific dependencies
• Protect proprietary knowledge. Verify release authority, flag customer-owned assets, and detect decision history and datasets that leak architecture or PII
• Change deployment region. Untangle region-specific endpoints, certificates, and identity providers before deploying elsewhere

Development workflow

• Clean a fork. Remove organization-specific changes before contributing upstream or publishing it
• Prepare a demo. Avoid exposing real data, tenant IDs, or environment configuration

Workflow stages

The plan is the shared record for decisions, approvals, changes, and residual risks across all three stages. When verify reports residual findings, run apply again to address them. Repeat until verify passes or only documented exclusions remain.

Plan mode goes beyond literal matches when the outcome calls for it. A multi-lens coupling review finds customer-specific logic, data models, and authorization patterns that survive a plain rename, and a modernization pass recommends portable alternatives for tenant-specific gateways, identity providers, and platform services. Findings outside the current run's scope route to the track that owns them (deployment, modernization, security) instead of being silently dropped.

Prerequisites

The target repository can use any language or framework. Mixed Python, .NET, Java, JavaScript, infrastructure, and documentation repositories are supported.

• Python 3.11 or later for the bundled scripts
• Git for repository discovery and delta scans
• GitHub Copilot, Claude Code, or another agent that discovers SKILL.md packages
• Gitleaks or TruffleHog on PATH when certifying Git history for public release

The Python scripts use only the standard library. No package installation is required.

Installation

Clone the repository and copy the skills/ directory into your project:

Claude Code:

git clone https://github.com/lshade/solution-debranding.git /tmp/solution-debranding
cp -r /tmp/solution-debranding/skills/* .claude/skills/

GitHub Copilot:

git clone https://github.com/lshade/solution-debranding.git /tmp/solution-debranding
cp -r /tmp/solution-debranding/skills/* .github/skills/

Other agents:

git clone https://github.com/lshade/solution-debranding.git /tmp/solution-debranding
cp -r /tmp/solution-debranding/skills/* <your-skills-directory>/

This installs four sibling directories:

skills/
   solution-debranding/
   solution-debranding-plan/
   solution-debranding-apply/
   solution-debranding-verify/

solution-debranding contains the shared framework and references used by the three stage-specific skills.

[!IMPORTANT] Install all four sibling directories without flattening them. The stage skills resolve the shared package by relative path.

Try it

Strip a brand and prepare for public release:

/solution-debranding-plan Contoso scope=full release-scope=public

Strip a brand and rebrand with a design system:

/solution-debranding-plan Contoso scope=full design-system=https://fluent2.microsoft.design/

Keep the current brand but adopt a design system:

/solution-debranding-plan Contoso scope=branding design-system=https://carbondesignsystem.com/

White-label for a new organization:

/solution-debranding-plan Contoso scope=full target=Northwind

Scan first and see what would need to change:

/solution-debranding-plan Contoso scope=full

Then apply and verify:

/solution-debranding-apply docs/debranding/2026-08-26-001-contoso-plan.md
/solution-debranding-verify docs/debranding/2026-08-26-001-contoso-plan.md

Generated evidence

Each run produces one plan file under ARTIFACT_ROOT (default: docs/debranding/). The plan carries the summary, implementation units, assessment, and validation evidence. See the full artifact format.

Safety constraints

• Plan never modifies product code
• Apply requires an existing plan and recorded approvals
• Apply changes one unit at a time
• Verify reports failures without fixing them
• Git history, credentials, and external systems require explicit authorization

Read the full safety contract.

Integration with other planning workflows

Solution Debranding is designed to coexist with plan-driven workflows like Compound Engineering. It treats planning artifacts from other skills (docs/plans/, docs/solutions/, docs/brainstorms/, etc.) as history records, scanning them for brand exposure without rewriting or deleting them. The plan phase asks you to decide the disposition of each planning directory (delete, anonymize, or exclude from the public tree) rather than silently dropping work that other workflows depend on.

FAQ

No. The scripts use only the standard library.

No. Plan mode scans and writes a plan artifact. It does not touch product, configuration, or infrastructure files.

Not automatically. The workflow reports copyright, ownership, attribution, bundled fonts, and licensed assets for review. It does not remove legitimate third-party names, licenses, standards, or required attribution. Confirm that you have authority to change or redistribute each item before approving it.

Any agent that discovers SKILL.md packages. The package avoids agent-specific APIs, but discovery and invocation syntax can differ between agents.

Yes. Pass design-system=https://fluent2.microsoft.design/ (or any structured system URL) to /solution-debranding-plan. The plan first identifies the real consumer and supported theming API, then maps only the values that integration requires. If the repository has no design-system consumer, the plan asks whether adoption is actually in scope instead of generating an unused token layer.

The scanner detects literals, opaque identifiers, and tenant-specific authorities. It does not detect coupling that survives renaming. Run the coupling review documented in the shared skill's references for semantic analysis.

Contributing

See CONTRIBUTING.md for validation commands and submission guidelines.

License

MIT. See LICENSE.