methodology-process

Michael Polanyi 的思维框架。基于7本核心著作深度调研，提炼6个心智模型、8条决策启发式。 用途：知识传承顾问，用 Polanyi 视角分析隐性知识传递、技能习得、科学哲学问题。 触发词：「用 Polanyi 的视角」「Polanyi 会怎么看」「隐性知识」「tacit knowledge」 「帮我用 Polanyi 的角度想想」「为什么说不清楚」「怎么传承经验」「知识管理」「师徒制」

Interactive workflow for auditing a SpecRoute artifact before it is treated as implementation-ready. Use when checking a PRD, spec triplet, prompt set, agent file, skill, command, hook, rule, or runtime layout. Asks which artifact path to inspect and which contract to apply.…

Execute implementation tasks with different approaches (single, dual-agent, plan-based).

Deep-dive investigation with documentation output.

Create implementation plans with phase management.

Code quality analysis through cleanup analysis and multi-perspective reviews.

Interactive task definition with Q&A and documentation.

Evaluate current session for skill-worthy workflows and create reusable skills. Triggers on: auto-skill, create skill from session, save workflow, capture this as a skill.

Detects stale project plans and suggests session commands. Triggers on: sync plan, update plan, check status, plan is stale, track progress, project planning.

Verify React component purity - components only render UI, all logic belongs in hooks. Use for: property drilling, logic in JSX, useEffect in component, domain useState in component, GUI and business logic separation.

Finalize guard for review, verification, anchoring, and metrics, including RAG/skills change branch.

Bootstrap skill for graph initialization and refresh. Use when a repository has no graph yet, has stale graph artifacts, or needs an initial build after project paths were discovered.

Interactive bootstrap skill for building a project-specific grooming roster. Use when you need to define which experts, reviewers, and domain specialists should participate in planning and validation for this repository.

Auto-trigger grooming based on complexity and risk signals, with strict micro-change skip criteria.

Core orchestrator rules for skill-first routing, performance, and workflow continuity.

Runner-first for PowerShell: every instruction is executed via .github/agents/invoke-isolated-powershell.ps1 (Mode detached by default), with try/catch/finally contract and auditable result.

Rules for writing reliable PowerShell with Set-StrictMode -Version Latest: mandatory @(), avoiding .Count on unknown type, safe null comparisons, variable interpolation, type conversion. Generate scripts adapted to current machine - detect Pester version, PSVersion, and…

Interactive bootstrap skill for preflight profiles. Use when you need to decide which preflight checks exist for a repository, which stacks they apply to, and what must block finalize or publish.

Interactive bootstrap skill for project-specific standards. Use when you need to capture coding standards, architecture rules, testing policy, language conventions, and repository-specific decisions with the user in the loop.

Practical anti-hallucination workflow for RAG: deterministic vs retrieval, query_type/workflow_phase routing, top-k/scoring, grounding/freshness/waste validation, and trends. Use for: decision answers, status/debug, and reducing retrieval hallucinations.

--- name: rag-domain-integrity-gate description: "Fail-fast gate and domain integrity repair in the RAG graph. Use for: path-like domain, domain noise, selective refresh safety, workflow-knowledge normalization." argument-hint: "Provide symptoms (e.g. domain_path_like>0),…

React performance optimization - memoization, list virtualization, lazy loading, bundle size. Use for: slow render, list with thousands of elements, excessive re-renders, useMemo, useCallback, React.memo, lazy, Suspense.

Read-only exception for single-command or single-script flows with strict no-write constraints.

Split code into smaller units and refactor while preserving DRY/KISS/SOLID. Use for: god file, large classes/components, too many responsibilities, too many constructor dependencies, domain split.

Control context budget and file-reading discipline. Use for: token reduction and scope minimization.