consensus-permission-escalation-guard

What This Skill Does

The consensus-permission-escalation-guard acts as an automated, deterministic governance layer for any OpenClaw agent workflow involving IAM or permission modification. Rather than allowing agents to modify system roles arbitrarily, this skill serves as a mandatory gatekeeper. It validates every escalation request against a strict schema, ensuring no malformed inputs bypass security logic. By evaluating requests against hard-coded policy flags, it identifies dangerous patterns—such as wildcard permissions, separation-of-duties violations, or missing incident references—before any change is enacted.

The skill operates in two primary modes: persona mode, which uses internal deterministic logic for standard authorization, and external_agent mode, which allows for complex multi-party voting and auditing. Upon evaluation, the skill returns an explicit ALLOW, BLOCK, or REQUIRE_REWRITE decision. Crucially, it enforces idempotency and generates auditable board artifacts for every transaction, ensuring that security teams have a complete trail of the decision-making process for post-incident analysis.

Installation

You can add this skill to your OpenClaw environment using the following command:

clawhub install openclaw/skills/skills/kaicianflone/consensus-permission-escalation-guard

Ensure that you have consensus-guard-core installed, as it is a required dependency for the aggregation and state management logic. The skill requires Node.js and tsx to execute the policy logic locally.

Use Cases

1. Automated IAM Provisioning: Preventing agents from granting 'AdministratorAccess' or wildcards to transient task roles.
2. Break-Glass Procedures: Ensuring that urgent production escalations are tied to a valid incident or ticket identifier before approval.
3. Separation of Duties (SoD): Validating that an agent requesting permission to create a user profile is not simultaneously the one approving the elevation of that user's rights.
4. Audit Compliance: Creating a structured record of why a specific elevation was allowed or denied, useful for SOC2 or internal security audits.

Example Prompts

1. "OpenClaw, please evaluate the proposed IAM update in ./policies/dev-to-prod.json and run the consensus-permission-escalation-guard to check for policy violations."
2. "Run the consensus-permission-escalation-guard on the current workflow. If the status is REQUIRE_REWRITE, suggest a more specific, non-wildcard policy definition for the requested role."
3. "Analyze the escalation request for the database migration task and use the external_agent mode to aggregate voting from the security_audit_team agent before applying changes."

Tips & Limitations

• Deterministic Nature: This skill is designed for safety, not flexibility. It will strictly block any request that does not perfectly adhere to the defined schema.
• Configuration: Ensure CONSENSUS_STATE_ROOT is defined in your environment to prevent state file write errors.
• Auditability: Periodically rotate or archive your consensus state files, as this skill performs continuous file writes to maintain the integrity of the audit logs.