Audit AI and cognitive systems against seven GMI intelligence invariants and efficiency principles.
npx skills add https://github.com/PatriotAi/ai-lab --skill gmi-audit-patriotai
Please help me install this Agent Skill. Skill: gmi-audit Source: https://github.com/PatriotAi/ai-lab/tree/main/.claude/skills/gmi-audit Command: npx skills add https://github.com/PatriotAi/ai-lab --skill gmi-audit-patriotai
High-intent search queries and answers about installing and using this skill.
Run a GMI audit that scores the system against seven invariants: world model, goal vector, inference, memory, learning, metacognition, and read-out. Each invariant is marked implemented, partial, or missing, with a prioritized list of gaps and improvements.
G1 is a predictive world model, G2 a value/goal vector, G3 inference and planning, G4 memory as attractor, G5 learning across sessions, G6 metacognition, and G7 determination or read-out. Generality is treated as an emergent property, not an invariant itself.
The G5 probe applies a cold-start recovery gate: the system needs external memory with self-contained critical references and a recovery checklist. Scores range from no recovery (red) to manual external loops (yellow) to automatic consolidation and retrieval cycles (green).
No. The audit produces a completeness map of intelligence invariants, not a VALID/INVALID verdict. For correctness or validity checking of an architecture or artifact, the skill explicitly delegates to a validation-mesh process instead.
They mark evidence quality: [E] for empirical findings, [C] for conceptual claims, and [S] for speculative ones. Conclusions should not rest on [S]-tagged items, and anything undetermined is left as UNKNOWN rather than fabricated.
aoecosystemValidate Mermaid diagrams, SoW coverage, and placeholders before architecture approval.
RynaroStore and recall task logs, patterns, and checkpoints in Markdown files.
Pabloprogramador23Compare requirements.md, roadmap.md, and actions.md for inconsistencies and coverage gaps.
yashs33244Benchmarks AI models across providers on latency tokens and cost.
