ghengis-skills

Use after completing a major work batch (plan execution, feature build, multi-task subagent run) to refresh code-graph-derived data in MEMORY.md/CONTEXT.md/SKILL_MEMORY, append dated narrative entries to TODO/CHANGELOG/lessons-learned, mine cross-project structural habits, and ratchet reusable permissions to global settings. Keeps project DNS in sync with reality.

Use when orchestrating multiple ghengis-skills across a single task -- coordinates pre-action validation, execution, post-action verification, and lesson extraction via a shared JSON scratchpad. TRIGGER when dispatching a subagent, finishing a major task, or when the user mentions "run the full chain", "validate this", "fire the pipeline", or explicitly invokes a chain by name.

Use to build and maintain an evolving understanding of user preferences, communication style, and working patterns — learns from interactions and adapts behavior over time

Use when starting a new project or organizing an existing one — creates self-documenting 4-layer project structure with MEMORY.md, CONTEXT.md, modular CLAUDE.md, and workspace routing

Use to accumulate and retrieve domain knowledge from past tasks — builds a grepable plain-text knowledge base that grows over time, enabling agents to learn from experience without vector databases

Use when reorganizing an existing project's file tree based on what actually connects to what — analyzes import graph, classifies layers, proposes an ideal target tree, validates the plan adversarially, then executes file moves with atomic per-language import rewrites. Always plan-first; never moves a file before the user confirms. TRIGGER on "reorganize this project", "clean up the file tree", "the structure is a mess", "make this match a real project layout", or after project-scaffold detects layout drift.

Use when the user wants to design a parametric 3D-printable part, mechanical bracket, enclosure, robot frame, sheet-metal panel, or any CAD model from a natural-language description. Calls the JARVIS build123d pipeline (Claude Max → build123d code → STEP/STL/3MF/DXF + topology sidecar) and verifies geometry via the bundled inspector. Cost is $0 when CLAUDE_MAX_ENABLED is set.

TRIGGER when user says "let's add a skill", "make a skill for X", "port the Y skill from superpowers", "I keep repeating this instruction", "this should be a skill", "create a skill"; OR when noticing a repeated multi-message instruction pattern that would benefit from being captured once. Use to create or edit any SKILL.md file in plugins/ghengis-skills/skills/.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work. Guides completion through structured options - merge locally, push and open PR, keep as-is, or discard. Verifies tests before presenting options; refuses to proceed if any test fails. TRIGGER on "let's wrap this up", "ship it", "merge this", "ready to ship", "done coding", or after build-validate chain ends with revised-and-shipped outcome.

Use when encountering ANY bug, test failure, error message, performance issue, or unexpected behavior, BEFORE proposing fixes. Forces root-cause investigation before any code change. Iron law - no fixes without evidence. TRIGGER on "this is broken", "test is failing", "why is X doing Y", "it crashed", "production issue", or any unexpected output. Cross-refs test-driven-development for writing the regression test, bug-hunt chain for the canonical fix pipeline.

Use BEFORE any creative or implementation work — creating features, building components, adding functionality, modifying behavior, or porting a skill. Explores user intent, requirements, and design through inline conversational dialogue. Never enters plan mode. Never uses structured question pickers. Inline chat only. TRIGGER when user says "let's build", "I want to make", "add a feature", "design X", or describes a new project; or BEFORE invoking any chain that builds a deliverable.

Use when implementing ANY feature, bugfix, refactor, or behavior change BEFORE writing implementation code. Forces RED-GREEN-REFACTOR - write the failing test first, run it to confirm it fails for the right reason, write minimal code to make it pass, commit, then refactor. If you didn't watch the test fail, you don't know if it tests the right thing. TRIGGER when starting any code change with testable behavior. Cross-refs systematic-debugging (Phase 3 writes the regression test), writing-skills (TDD on documentation), feature-build/bug-hunt chains.

Run once after installing ghengis-skills to configure autonomous permissions — sets up safe tool access and blocks dangerous operations

One-shot installer for the agent-monitor terminal status line. Copies statusline-bar.py to ~/.claude/ and adds the statusLine config to settings.json. Run once per machine. Idempotent and safe.

Force-refresh the ghengis-skills plugin from GitHub, bypassing Claude Code's `/plugin update` caching. Use when `/plugin update` claims you're on an old version even after running it, or when you want to guarantee you're on the latest. Idempotent and safe.

Real-time subagent monitoring dashboard and status line -- tracks agent spawning, completion, permissions, and history via hooks. Auto-opens a browser dashboard when agents are active. Includes a terminal status line showing model name and context window usage.

Use when building autonomous workflows that need accountability -- provides an immutable hash-chained audit trail pattern for tracking what agents did, when, and why, with tamper detection

Use after any agent or subagent claims a task is complete -- verifies completion through signal detection, catching premature "done" claims, partial outputs, and unfinished work before it ships

Use when reviewing agent output for factual accuracy -- detects fabricated URLs, unsourced statistics, fake citations, impossible future claims, and confident assertions without evidence

Use when dispatching subagents or writing role-specific instructions -- provides 22 specialized role templates for researcher, builder, validator, and 19 other roles

Use when writing, reviewing, or improving prompts for agents, subagents, or LLM calls — detects anti-patterns, suggests fixes, and enforces prompt quality principles

Give Claude a sense of time -- tracks elapsed time between messages, task durations, project switching, and activity patterns. Hooks inject temporal context into every prompt automatically. Includes a portable Python module for wrapping any LLM call with time awareness.

Use when the user shares a research paper, technical doc, engineering blog post, or asks how to apply external knowledge to their codebase -- covers strategic reading, contribution mapping, spec delta writing, and the "what NOT to adopt" framing. TRIGGER when message contains arxiv.org, a paper/doc URL, "this paper", "read this", "apply this research", "how do we use this", or references a technique from an external source (Anthropic docs, OpenAI blog, conference paper, engineering post).

Use when building agents or workflows that should learn from their own measurable outcomes -- covers fitness signal design, cognition store schema, Analyzer prompts, UCB1 retrieval, poison mitigations, and audit loops. TRIGGER when user mentions "learn from outcomes", "agent memory", "feedback loop", "lesson store", "cognition", or "agents getting smarter over time".

Use when starting any conversation - establishes how to find and use ghengis-skills, requiring Skill tool invocation before ANY response when a ghengis-skill applies (mirrors superpowers:using-superpowers pattern)

Use when a task benefits from multiple creative perspectives -- spawns parallel agents with different viewpoints, then synthesizes the best elements into a final output

Use during long-running sessions or complex multi-step tasks -- detects context window degradation, task drift from original request, and offers recovery strategies including checkpoint restart and re-anchoring

Use when a task is too large for a single session — provides multi-session execution with checkpoints, progress tracking, and resume capability so complex projects survive context resets and session boundaries

Use when facing complex tasks that need multiple specialized agents working in sequence or parallel — task decomposition, role assignment, dependency DAGs, revision loops, and quality validation

Use when helping with financial record-keeping, expense categorization, tax preparation, or accounting questions — covers double-entry accounting, IRS categories, reconciliation, and financial reports

Use when analyzing codebases, understanding architecture, tracing dependencies, or building mental models of unfamiliar code -- covers AST parsing patterns, 6-layer architectural classification, import graph analysis, and structural code search

Use when creating written content -- blog posts, documentation, marketing copy, technical writing -- covers structure, audience targeting, SEO basics, and editorial quality

Use when managing client relationships, freelance projects, or any work involving multiple clients — covers client tracking, project lifecycle, communication logging, pipeline management, and relationship health

Use when analyzing data, creating visualizations, or extracting insights -- covers statistical methodology, pandas patterns, data cleaning, visualization best practices, and insight presentation

Use when a research task requires iterative depth -- goes beyond single-pass research with a 7-phase methodology that includes gap analysis, targeted follow-up, adversarial review, and convergence checking

Use when setting up deployment, CI/CD, Docker, or infrastructure -- covers solo-dev deployment patterns, GitHub Actions, Docker best practices, SSL, environment management, and rollback procedures

Use when working with ESP32 firmware or IoT hardware projects — covers PlatformIO build system, I2S audio, WiFi/BLE patterns, memory management, and state machine design

Use when building or modifying FastAPI applications — covers async patterns, Pydantic v2 models, dependency injection, WebSocket auth, middleware, and blocking code handling

Use when helping organize files, clean up directories, or establish file naming and structure conventions — covers 18 manifest types, intelligent file categorization, naming conventions, and organizational patterns with audit trails

Use when building or modifying Flutter/Dart applications — covers widget patterns, state management, platform channels, navigation, and common Dart conventions