Local-first AI jury skill: 9 seats deliberate, 10 scoring functions audit claims, human final.
SaferSkills independently audited ai-judge (Agent Skill) and scored it 92/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Installs the AI Judge multi-model deliberation system as a Hermes-compatible skill. Once installed, you can:
ai-judge jury --question "..." — Create a jury session with 9 AI seatsai-judge collect --run latest — Collect independent answers from all seatsai-judge verdict --run latest — Generate auditable verdict with evidence scoringai-judge reflect --date YYYY-MM-DD — Daily performance reflectionai-judge score-v2 --demo — Run the public v2 scoring-engine demoai-judge neuro-profile --demo — Run the v3.1 cognitive proxy demoai-judge hard-truth --demo — Preview L0-L4 judgment-first feedbackai-judge v3-pipeline --demo — Run the full v3.1 pipelineai-judge v3.2-pipeline --demo — Run evidence + dissent + reasoning-tree demoai-judge seats --list — Inspect the 9 fixed COUNCIL-004 persona seatsai-judge trace --demo — Detect shared citation sources across model seats# Install via pip
pip install -e .
# Or via Docker
docker pull ghcr.io/reguorier/ai-judge:latestai-judge license status
ai-judge --help
ai-judge v3-pipeline --demo
ai-judge v3.2-pipeline --demo
ai-judge seats --list
ai-judge trace --demoThis skill package contains the public open-core layer:
| Included (Public) | Not Included (Paid Core) |
|---|---|
| CLI surface, packaging, and v2/v3/v3.2/v3.3 demos | Production collector engine |
| v2 formula engine, v3 cognitive proxy demos, v3.2 evidence/dissent demo, COUNCIL-004 persona/trace tools | Production browser/CDP orchestration |
| macOS bridge source code | Managed scoring service/runtime |
| Public documentation and schemas | SaaS license server |
| TypeScript UI components and Rust reference engine | Managed desktop app distribution |
| Docker entrypoint and compose | Team/enterprise integrations |
| BSL 1.1 license terms | Support/SLA layer |
| Prompt templates | Hosted deployment tooling |
Production commands (jury, collect, verdict, reflect) require the paid ai-judge-core package.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.