wardrobe

The Pat Helland lens: treat durable data as immutable, append-only facts that carry their own identity and version; and design across service, machine, or org boundaries assuming you cannot have distributed transactions — only idempotent, at-least-once messages that must be reconciled. Apply this on in-place updates and overwrites, "just UPDATE the row," two-phase commit or distributed transactions at scale, exactly-once delivery assumptions, shared mutable state across services, and any data that crosses a boundary of service, machine, org, or time. Trigger even when Helland isn't named — the smell is mutation-in-place and pretending distributed systems can be transactional.

The Rich Hickey lens: pursue simple (un-braided, one role) over easy (familiar, near-at-hand), decomplect concerns that got tangled together, and separate value, identity, state, and time. Apply this whenever "easy," "convenient," or "it's what we already use" is doing the work of justifying a design; when state is braided into logic, policy into mechanism, or data into behavior; when object-soup would be plainer as data; and when a hard design decision is being made at the keyboard under time pressure instead of thought through first. Trigger even when Hickey isn't named — the smell is complecting, and confusing "easy" with "simple."

The John Carmack lens: minimize the mutable state you have to reason about, do the simple thing before the clever one, and treat correctness tooling — static analysis, warnings-as-errors, determinism — as non-negotiable. Apply this on sprawling mutable state, "we'll clean it up later," debugging by print- statement whack-a-mole, silenced compiler warnings, nondeterministic or non-reproducible behavior, premature abstraction, perf work that skips measurement, and any "what should we actually work on" prioritization call. Trigger even when Carmack isn't named — the smell is code whose full set of states you can't hold in your head, and effort that isn't ruthlessly focused.

The Fabrice Bellard lens: build the smallest self-contained thing that fully does the job, get performance from insight instead of hardware, and reuse existing machinery instead of reinventing it. Apply this whenever code is growing dependencies, frameworks, services, or infrastructure to solve a problem that a smaller, sharper implementation could handle — or whenever someone says a thing needs a cluster, a big framework, or a rewrite. Use for size/complexity audits, "do we really need this dependency/service?" questions, performance work where the reflex is to add hardware or caching, green-field "should this be one file or a platform?" decisions, and any moment the answer is trending toward "that's impractical, you'd need X." Trigger even when the user doesn't name Bellard — the smell is scope inflation and reaching for resources over insight.

Use when iteratively researching a topic and filing the synthesis into an Obsidian vault for later ingestion. Output goes to `<vault>/raw/` (NOT directly into wiki pages) so that `vault-ingest` can process it on its own schedule. Triggers: '/vault-autoresearch', 'autoresearch [topic]', 'research [topic] into the vault', 'deep dive into [topic]', 'find everything about [topic] for the vault', 'research and stage'.

Ingest sources into the Obsidian wiki vault. Reads a source, extracts entities and concepts, creates or updates wiki pages, cross-references, and logs the operation. Supports files, URLs, and batch mode. Triggers on: ingest, process this source, add this to the wiki, read and file this, batch ingest, ingest all of these, ingest this url.

Health check the Obsidian wiki vault. Finds orphan pages, dead wikilinks, stale claims, missing cross-references, frontmatter gaps, empty sections, misfiled pages, and catalog gaps (pages absent from a bucket _index.md or the master index.md). Creates or updates Dataview dashboards. Generates canvas maps. Triggers on: "lint", "health check", "clean up wiki", "check the wiki", "wiki maintenance", "find orphans", "wiki audit".

Claude + Obsidian knowledge companion. Sets up a persistent wiki vault, scaffolds structure from a one-sentence description, and routes to specialized sub-skills. Use for setup, scaffolding, cross-project referencing, and hot cache management. Triggers on: "set up wiki", "scaffold vault", "create knowledge base", "/wiki", "wiki setup", "obsidian vault", "knowledge base", "second brain setup", "running notetaker", "persistent memory", "llm wiki".

Fast BM25 (lexical-only) retrieval over the Flight Planner wiki via qkb. No LLM, no graph, no rerank — cheap and deterministic. Use for known-term questions (proper nouns, acronyms, exact phrases), AB-test baselines against the full pipeline, and quick smoke checks. Triggers on: vault-query-bm25, bm25 vault query, lexical vault search, fast keyword search the wiki, qkb bm25, search the wiki without LLM.

Pure graph-traversal retrieval over the Flight Planner Obsidian wiki vault. Uses qkb's graph layer directly (LINKS_TO / EMBEDS / REFERENCES edges across 1800+ typed nodes) — no BM25, no vector, no LLM expansion, no rerank. Use for entity-relation questions ("what does X link to / depend on / feed"), structural mapping ("what's connected to Y"), and following established wikilinks across the corpus. Triggers on: vault-query-graph, graph vault query, what does X link to, what feeds X, what depends on X, how does X relate to Y, graph traversal, walk the graph from X.

Answer questions using the Obsidian wiki vault. Default skill for "what do you know about", "explain", "summarize", "find in wiki", "search the wiki". Two-mode router: when qkb is installed and not opted out, escalates from curated read (hot.md + index.md + page reads) to qkb's BM25 or graph primitives on questions the curated path can't reach. When qkb is opted out (QKB_DISABLED=1) or not on PATH, runs as pure curated read with honest gap reporting. Supports quick, standard, deep modes; files answers back as wiki pages so knowledge compounds. Triggers on: query:, what is, explain, summarize, find in wiki, search the wiki, based on the wiki, wiki query quick, wiki query deep.

Save the current conversation, answer, or insight into the Obsidian wiki vault as a structured note. Analyzes the chat, determines the right note type, creates frontmatter, files it in the correct wiki folder, and updates index, log, and hot cache. Triggers on: "save this", "save that answer", "/save", "file this", "save to wiki", "save this session", "file this conversation", "keep this", "save this analysis", "add this to the wiki".

Publish a self-contained HTML file to a shareable encrypted Cloudflare-hosted artifact URL using html-artifact-publisher. Use when asked to upload, publish, share, or host a generated HTML page/artifact/explainer via artifacts.craftdesign.group or the HTML artifact publisher.

Use this skill whenever the user needs to choose the correct UI component for forms, notifications, errors, loading, buttons, overflow, onboarding, or design system contributions. Always start with the Main Entry Point dispatcher to select the right tree, then follow the exact branching logic. This skill prevents ad-hoc component choices by enforcing production design system decision trees.

Use during SPDD design review before generation to validate entities, responsibilities, interfaces, boundaries, dependencies, and task granularity.

Use before SPDD analysis or implementation to lock business intent, domain language, scope boundaries, acceptance criteria, dependencies, and hard constraints.

Use after SPDD alignment to analyze requirements against relevant code, domain concepts, risks, gaps, strategy, and acceptance coverage before writing a REASONS Canvas.

Use in SPDD when API or system-boundary behavior must be verified with executable scenarios covering normal, boundary, and error cases from the REASONS Canvas.

Use when implementing code from an approved SPDD REASONS Canvas, following Operations in order and staying inside the canvas Norms and Safeguards.

Use after SPDD-generated changes to review behavior, prompt/code consistency, architecture boundaries, code quality, and whether to update prompt first or sync prompt after refactor.

Use in SPDD when requirements, behavior, acceptance criteria, business rules, or safeguards change; update the REASONS Canvas before changing code.

Use to create or review a Structured-Prompt-Driven Development REASONS Canvas before code generation: Requirements, Entities, Approach, Structure, Operations, Norms, Safeguards.

Use when a large SPDD requirement needs to be split into independent, deliverable user stories before analysis and REASONS Canvas work.

Use after behavior-preserving refactors or implementation cleanup in SPDD to synchronize the REASONS Canvas back to the current code without changing business intent.

Use when a session is working under Structured-Prompt-Driven Development (SPDD), when prompts/specs must be versioned delivery artifacts, or when logic-heavy changes need a prompt-first closed loop.

Format a status update, progress report, summary of work, or explanation of how concerns are being addressed — in plain prose suitable for handoff to a non-technical stakeholder. Use whenever the user asks for a status update, an explanation of work, a summary of what's been done, or how something is being addressed. The output should read like a confident pilot-engineer wrote it, not an AI.

Use when assessing, designing, fixing, or explaining Continuous Delivery in Dave Farley's style. Triggers on continuous delivery, CD, CI/CD maturity, deployment pipelines, release readiness, trunk-based development, release process audits, DORA metrics, "can we deploy now?", "Farley", or requests to make software releasable safely and frequently. Also use for regulated or legacy systems where the user needs a pragmatic CD adoption plan.

Use when the user says "delegate to subagent", "use cavecrew", "spawn investigator/builder/reviewer", "save context", or "compressed agent output". Decision guide for delegating to caveman-style subagents (`cavecrew-investigator`, `cavecrew-builder`, `cavecrew-reviewer`) whose tool-results return ~60% smaller than vanilla equivalents, so the main context lasts longer across long sessions.

Use when the user invokes /caveman-stats or asks for caveman token-usage stats. Shows real token usage and estimated savings for the current session, read directly from the Claude Code session log — no AI estimation. Output is injected by the `caveman-mode-tracker` hook (vendored alongside this skill), which intercepts the slash command and returns the formatted stats as a blocked-decision reason.

Drive an issue from triage to merged PR with minimal operator intervention. Runs the full loop — read & label, post agent brief, write a failing test, implement minimally, self-review via subagent, replicate local CI, push, watch CI, squash-merge, clean up, and report with next-step suggestions. Use when the operator says "ship #X", "do issue #X end-to-end", "take #X to merge", "fix and ship #X", "do #X autonomously", or any phrasing that authorizes carrying a single issue all the way through. Also use when the operator follows a /triage call with "do as you'd like", "update the issue then start implementation", or otherwise hands off the rest of the workflow. Do NOT use for purely advisory questions ("what should we do about #X?", "look at #X with me") — those stay in regular triage / discussion mode.

Use when the user says "write a commit", "commit message", "generate commit", "/commit", or invokes /caveman-commit, or when staging changes. Generates ultra-compressed Conventional Commits messages — subject ≤50 chars, body only when "why" isn't obvious. Cuts noise while preserving intent.

Use when the user invokes /caveman-compress FILEPATH, says "compress memory file", or otherwise asks to compress natural-language memory files (CLAUDE.md, todos, preferences). Rewrites the file into caveman format to save input tokens, preserving all technical substance, code, URLs, and structure. The compressed version overwrites the original; a human-readable backup is saved as FILE.original.md.

Use when the user invokes /caveman-help, says "caveman help", asks "what caveman commands", or asks "how do I use caveman". Displays a one-shot quick-reference card for all caveman modes, skills, and commands. Does not change mode or persist anything.

Use when the user says "review this PR", "code review", "review the diff", "/review", or invokes /caveman-review, or when reviewing pull requests. Generates ultra-compressed code review comments — one line per finding: location, problem, fix. Cuts noise while preserving actionable signal.

Use when the user wants ultra-compressed responses — says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra.

Use when reviewing or refactoring rtk Rust code that feels verbose, over-engineered, or non-idiomatic — applies iterator chains, early returns, Option/Result chaining, and zero-copy patterns while leaving rtk's mandatory constraints (lazy_static regex, .context() chains, fallback arms) intact

Use when designing a new rtk filter module or refactoring an existing one — picks between Newtype, Builder, State Machine, Trait Object, RAII, Strategy, and Extension Trait, with rtk-flavored guardrails against premature abstraction

Use when auditing open GitHub issues on the rtk repo (or any project) — categorises by Bug/Feature/Enhancement/Question, scores risk red/yellow/green, detects duplicates via Jaccard similarity, cross-references open PRs (`fixes #N`), flags stale issues, runs deep analysis via parallel agents on demand, and posts validated draft comments. Args: `all` for deep on every issue, issue numbers to focus, `en`/`fr` for language.

Use when investigating, benchmarking, or fixing rtk performance regressions — enforces <10ms startup, <5MB resident memory, 60-90% token savings, <5MB stripped binary; covers lazy_static regex, zero-copy parsing, dependency minimisation, and flamegraph workflow

Use when batch-reviewing rtk PRs from simplest to most complex — checks each PR's state (conflicts, CLA, CHANGES_REQUESTED), reads the full diff, presents one PR at a time with a 2-4 sentence summary and explicit merge prompt, never merges without a typed "ok", and posts boldguy-tone comments on blocked PRs (rebase needed, CLA missing). Args: `triage` to chain `/rtk-triage` first; `from:<num>` to resume.