Consensus Publish Guard
PassAudited by ClawScan on May 1, 2026.
Overview
This is a coherent local pre-publication review guard, but users should notice that it writes persistent audit artifacts and relies on npm dependencies.
This skill appears safe and purpose-aligned for local pre-publication review. Before installing, make sure you are comfortable with npm dependency installation, configure the consensus state path to a protected location, and avoid treating external vote text as trusted instructions.
Findings (3)
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.
A dependency update could affect guard behavior or state handling, even though this is normal for npm packages.
The skill relies on external npm packages and uses caret semver ranges, so compatible newer dependency versions may be installed depending on the install method.
"dependencies": { "consensus-guard-core": "^1.1.15", "tsx": "^4.20.3" }Use a locked installation workflow where possible and review the consensus-guard-core dependency if strict reproducibility is important.
Draft-derived review results, rewrite patches, or external vote text may remain in local board state after the check completes.
The skill writes persistent decision artifacts, including votes and response data, to the configured state path.
await writeArtifact(board_id,'decision',{ idempotency_key: idem, decision_id, final_decision, votes, aggregation, response },statePath)Store the consensus state path somewhere access-controlled, and avoid submitting secrets in drafts unless that audit trail is intended.
Untrusted or poorly attributed external votes could influence the final decision or be stored in the audit trail.
The skill can consume inputs produced by other agents or humans, which creates an inter-agent data boundary.
`mode="external_agent"`: caller supplies `external_votes[]` from agents/humans/models for deterministic aggregation.
Only pass external votes from sources the workflow trusts, and treat free-form vote reasons as data rather than instructions.