V19 Causal Auditor
ReviewAudited by ClawScan on May 10, 2026.
Overview
This instruction-only skill is broadly aligned with causal auditing, but it sends decision data to an unverified remote endpoint and describes automatic escalation plus irreversible audit logging without clear user control.
Before installing or using this skill, verify who operates the Cloudflare endpoint, avoid sending sensitive decision data, and require explicit approval before any arbitration submission or immutable audit-chain write.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Potentially sensitive decision context could leave the user's environment and be processed by an external service whose trust and retention practices are not documented here.
The skill directs posting decision identifiers, context, and options to a remote governance endpoint, but the artifacts do not define the service operator, data boundaries, retention, or identity verification.
curl -s -X POST https://boat-atlas-spa-flexible.trycloudflare.com/governance/causal-trace ... "decision_id" ... "context" ... "options"
Use only non-sensitive examples unless the service operator is verified; require explicit user approval before sending real decision data to the endpoint.
Incorrect or sensitive audit results could become permanent or difficult to retract.
The skill states that audit results are written to an immutable audit chain, creating persistent records without explaining retention, deletion, correction, or user review controls.
审计结果写入V89审计链，不可篡改
Add a clear confirmation step before persistent writes, document what is stored, and provide correction or deletion guidance where possible.
A mistaken audit could trigger further arbitration or governance actions without the user realizing it.
The workflow says high conflict scores are automatically submitted to another arbitration process, so a bad input or false conflict score could propagate to downstream governance without containment.
冲突评分超过阈值 → 自动提交V57仲裁
Require user confirmation before arbitration submission and document thresholds, rollback options, and the scope of downstream effects.
Users may not notice that the workflow involves a service credential and delegated access to the remote governance API.
The skill uses a governance key for the remote service. This is expected for an API integration, but the registry declares no primary credential or required environment variable.
-H "X-Governance-Key: <你的专属密钥>"
Declare the credential requirement clearly and advise users not to paste private keys into shared logs or conversations.
Users must trust the remote service implementation even though it is not included in the skill artifacts.
The skill's functionality depends on a remote Cloudflare-hosted governance service, while the package itself has no code for review.
https://boat-atlas-spa-flexible.trycloudflare.com/governance/register
Verify the service provenance and operator before using it for real governance or business decisions.