pai_skill

Overview

🤖 This skill presents PAI (Personal AI Infrastructure) as a core, authoritative reference for how the system magnifies human capabilities. It explains the PAI Algorithm, depth selection, and the operational phases that ensure rigorous, auditable problem solving. The description emphasizes structure, repeatability, and defense-in-depth for agentic workflows.

How this skill works

PAI inspects requests and enforces a mandatory algorithmic workflow that always runs and classifies effort into three depth levels: FULL, ITERATION, and MINIMAL. FULL mode expands into seven phases (Observe, Think, Plan, Build, Execute, Verify, Learn), creates binary ISC criteria, selects capabilities via a two-pass process, and records verification evidence. Hooks and thinking-tool assessments guide capability choice while ISC criteria drive measurable outcomes.

When to use it

• Designing or describing agentic AI workflows and governance
• Implementing or auditing task automation that requires verifiable outcomes
• Creating reproducible problem-solving pipelines for engineering teams
• Documenting decision processes where auditability and verification matter
• Building multi-agent compositions that require explicit capability selection

Best practices

• Always define ISC criteria as binary, testable, and exactly eight words
• Treat the FormatReminder hook as authoritative for initial depth classification
• Perform the two-pass capability selection: hook hints then THINK validation
• Justify exclusion for every thinking tool; don't skip without a reason
• Capture verification evidence for each ISC task before learning phase

Example use cases

• Authoring a reproducible agent workflow for system design using PAI phases
• Auditing automated task execution to ensure each ISC passes verification
• Teaching teams how to compose capabilities with named patterns (Pipeline, Gate)
• Converting ad-hoc problem solving into an auditable seven-phase process
• Designing a multi-skill agent that selects Research→Architect→Engineer