Find a skill to use

Reusable strategy for bridging two already-formalized abstraction layers in a proof assistant project.

Use when a user wants to analyze local AI assistant session history for concepts, terms, jargon, APIs, or mental models they asked AI to explain, then produce a privacy-safe knowledge-gap digest, review plan, or recurring learning backlog without sharing raw transcripts.

Use when a user wants to analyze local AI coding assistant session patterns and turn aggregate focus, friction, and repeated workflows into SkillFab skill request drafts without sharing raw transcripts.

Build polished, self-contained interactive HTML/SVG diagrams from nodes, edges, views, descriptions, and links. Use when Codex needs to create a clickable architecture diagram, flowchart, system map, research map, or technical-blog figure that works in a browser without a frontend framework.

Design, implement, and debug publication-ready interactive research architecture maps for technical blogs and research notes. Use when Codex needs to organize related papers by mathematical role, turn a static architecture diagram into clickable HTML/SVG, map citations onto a non-serial topology, distinguish pipeline factorization from CNN/Transformer backbones and diffusion/flow inference semantics, or integrate and validate such a figure in Markdown, Jekyll, or GitHub Pages.

Use when preparing, executing, observing, or diagnosing an OS/kernel boot on physical development hardware through firmware, a bootloader, or serial transport, especially when emulator evidence, artifact identity, recovery availability, one-shot boot commands, or early-boot silence must be handled safely.

Use when booting or debugging an OS/kernel under QEMU — silent hangs with no console output, panic loops that look like silence, distinguishing "slow" from "hung", hosts without QEMU binaries installed, or when choosing launch flags, serial/monitor setup, and kernel cmdline logging for a QEMU guest.

How to manage a Docker-based development environment well: reuse cached images instead of re-downloading toolchains, keep build containers alive, mount the right directory, and know when to commit vs rebuild. Use when starting any containerized build/test, debugging "container compiled the wrong code", "ELF/file not found in container", "stale build artifacts", "downloads toolchain every time", or when a repo has multiple worktrees and you must mount the correct checkout. Triggers: - Starting a containerized build or test (long-running or not) - Container can't see files that exist on the host - Build outputs look stale or come from the wrong branch - Every build re-downloads toolchains/dependencies - `docker commit` decision: save a configured environment or rebuild - Multiple checkouts (git worktrees) and choosing what to mount - Container exited and the background job vanished

Use when using QEMU for OS/kernel research or debugging — architecture bring-up experiments (e.g. RISC-V), MMU/memory-model behavior tests (Svade/Svadu, svpbmt, SMP), hardware-vs-emulator comparison (DTB, board quirks), reproducible experiments, emulator-based CI/regression, or debugging silent hangs, panic loops, and hosts without QEMU binaries (Docker fallback).

Use when the user wants to analyze local AI coding assistant session data (Claude Code, Codex, Kimi Code). Triggers on requests like "analyze my coding sessions", "what tools do I use most", "compare Claude Code vs Codex usage", "give me a summary of what I've been working on", or similar session-analysis inquiries.