consensus-persona-generator

What This Skill Does

The consensus-persona-generator is the essential foundation for building reliable multi-agent decision architectures within the Consensus.Tools ecosystem. Unlike standard agent pipelines where a single model critiques its own output—creating a loop of confirmation bias—this skill facilitates the creation of a structured "disagreement layer."

At its core, this skill creates N distinct, persistent decision personas. Each persona is engineered with specific traits including role-based perspectives, varying bias profiles, distinct risk postures, and unique voting styles. Once generated, these personas are bundled into a versioned persona_set artifact, which is then persisted directly to your board state. By standardizing these cohorts, the generator ensures that downstream guards and arbitrators have a consistent, diverse set of "voices" to review and approve task outputs. It handles the initial reputation baselines for these personas, delegating ongoing performance-based reputation updates to the complementary consensus-persona-engine skill, thereby maintaining a clear separation of concerns between initialization and evolution.

Installation

To integrate this skill into your local OpenClaw environment, ensure you have the necessary runtime dependencies (node and tsx) installed. Use the following command in your terminal:

clawhub install openclaw/skills/skills/kaicianflone/consensus-persona-generator

Once installed, you can trigger the generator directly using the command line interface to produce your initial artifact sets:

node --import tsx run.js --input ./examples/persona-input.json

Ensure your local environment configuration points to the correct consensus state path so that artifacts are persisted correctly within your board directory.

Use Cases

• Automated Governance: Bootstrap a committee of 5-7 specialized personas to act as an internal audit board for automated email responses or code deployment approvals.
• Risk Mitigation: Generate a persona cohort with high-risk aversion profiles to review content generation before publishing to social media or public-facing documentation.
• Diverse Feedback Loops: Create a reusable library of evaluators that represent different stakeholders (e.g., technical architect, end-user, security officer) to validate feature specifications.
• Research Validation: Use the persona sets to stress-test data analysis results by applying multiple interpretive frameworks to the same dataset.

Example Prompts

1. "Generate a new persona set for a security-focused audit of my current project, requiring at least 4 distinct risk-averse personas, and save it to the board as 'security-review-group'."
2. "Create a diversified cohort of 6 personas to evaluate my upcoming technical documentation; ensure the set includes at least one persona specializing in UX and one in backend infrastructure."
3. "Refresh the personas for the 'content-approval' board, replacing the existing set with a new cohort that focuses on creative, non-traditional perspectives to reduce editorial groupthink."

Tips & Limitations

• Reusability First: Before generating new personas, check for existing, compatible persona_set artifacts on your board. Reusing sets reduces state bloat and ensures long-term consistency in your decision-making metrics.
• Local Execution: By default, this tool is deterministic and local. It does not require network access unless you modify the backend to call external LLM APIs. Keep the default configuration to ensure high performance and privacy.
• Separation of Logic: Do not attempt to use this tool for adjusting persona reputation. Use the consensus-persona-engine for updating how personas behave based on their previous performance; let this generator handle the creation and structure only.
• Security: While the tool does not perform arbitrary code execution, always inspect your persona-input.json files when sourcing them from external repositories to ensure they meet your board's policy requirements.