Skill Registry

Use when the user explicitly wants to turn clean, human prose INTO maximal AI slop — as a joke, a demo of what AI tells look like, or to stress-test the humanize skill by feeding it the worst input imaginable. The evil twin of humanize — it adds every tell humanize strips. For laughs and demos only; never run it on a real deliverable.

Vast.ai Python SDK — high-level API for GPU instances, volumes, serverless endpoints, and billing.

Challenge, push back, play devil's advocate on AI output. Use when: challenge this, are you sure, push back, prove it, what if you're wrong, devil's advocate, stress test, poke holes, second opinion, sanity check, too confident, really?, question this decision. Subcommands: anchor (committed too fast), verify (facts wrong?), framing (wrong problem?), deep (full devil's advocate in separate context).

Sense-making before action. Classify problem using Cynefin triangulation (3 tests + decomposition) to route to the right skill chain. Use when: frame, what approach, how should I start, which skill, where to begin, unsure what to do. NOT for known tasks — just do them.

Query and display available GitHub Copilot AI models with their capabilities, context limits, and features. Use when: "what models are available", "show copilot models", "list github models", "check model capabilities", "switch models". Examples: - user: "What models can I use with GitHub Copilot?" → fetch and display available models - user: "Show me models with vision support" → filter models by capability - user: "Which model has the largest context window?" → compare model specifications - user: "List all GPT-5 models" → filter by model family

Converts a PRD or requirements document into a structured, phased implementation plan with individual phase files and granular per-task files written to .context/plans/. Also restructures existing monolithic planning documents into digestible, hierarchical directory structures. Creates a root plan index summarising all phases, a numbered phase file per phase, and a numbered task file per task inside each phase directory. Use when the user asks to create an implementation plan, break down a PRD, convert requirements to tasks, structure project phases, generate a roadmap, plan a project in sprints, organise task breakdown, split a monolithic planning doc, or decompose a spec into phases and tasks.

Resume session from CONTEXT-llm.md. Use when resuming work, loading saved context, continuing a previous session. Triggers include "load context", "resume session", "continue where I left off".

Recommend optimal Claude model (haiku/sonnet/opus) for a task. Use when user asks "which model", "pick model", "model for", or before starting costly/complex tasks. Covers tech and non-tech tasks.

Splits a large feature branch into smaller, focused pull requests using stacked branches and cherry-pick. Groups commits by concern (infrastructure, application logic, tests, housekeeping), proposes descriptive branch names for user approval, creates stacked branches, and generates What/Why MR titles and descriptions. Use when the user says a PR is too big, asks to split a PR, wants to decompose a branch, or needs to break work into reviewable chunks.

Safe-to-fail experiment for Complex domain problems where cause-effect is only visible in retrospect. Two-phase: foreground qualify → background probe → sense result. Use when: probe, safe-to-fail, test hypothesis, experiment with hypothesis, Complex domain with hypothesis. NOT for brainstorming (use brainstorm) or known cause-effect (use investigate).

Comprehensive toolkit for validating, optimizing, and understanding Prometheus Query Language (PromQL) queries. Use this skill when working with PromQL queries, prometheus queries, metrics queries, alerting rules, recording rules, or grafana dashboards to check syntax, detect anti-patterns, identify optimization opportunities, and interactively plan queries with users.

Analyze collaboration patterns (HOW) and compute metrics from captured sessions. Use when reviewing collaboration quality, analyzing human-AI interaction, computing session metrics. Triggers include "retrospect collab", "collaboration analysis", "session patterns", "how am I collaborating".

Analyze domain insights (WHAT/WHY learned) from captured sessions. Use when reviewing learnings, extracting patterns, analyzing decisions. Triggers include "retrospect domain", "domain analysis", "what did I learn", "session insights".

Save session to CONTEXT-llm.md with conversation summary. Use when saving work, checkpointing progress, preserving session state. Triggers include "save context", "save session", "checkpoint", "save my progress".

Use when the user asks to keep going, fix until green, converge on a verifiable target, or work within an explicit budget.

Use when plan/design is requested, whenever Codex is in Plan mode, or a non-trivial change needs evidence-backed scope, verification, or acceptance before action.

Use when user reports an error, bug, or something not working. Search-first troubleshooting with diagnostic phase. Triggers: debug, error, broken, not working, failing, crash, exception.

Groups plan phases and tasks into dependency-ordered waves for parallel subagent execution via git worktrees. Builds a task dependency DAG, assigns wave numbers via topological sort, emits a living wave document that tracks status as work lands, and updates wave progress when tasks complete. Use when asked to: group tasks into waves, plan parallel execution, schedule worktrees, create a wave breakdown, wave planning, dependency grouping, update wave statuses, parallel subagents, which tasks can run in parallel.

Discover, install, compare, and manage Claude Code skills. Search the registry, get recommendations, validate skill quality, and manage your installed skills.

Use when the user wants a new git worktree created for a task. Picks a kebab-case branch name with a fix/feat/chore/docs/refactor prefix, runs `git worktree add` from the configured base branch, and reports the new path.

Use when the user wants a new git worktree created for a task. Picks a kebab-case branch name with a fix/feat/chore/docs/refactor prefix, runs `git worktree add` from the configured base branch, and reports the new path.

Autoresearch loop for Claude Code skills — greedy keep/discard hill climbing on a 10-dimension quality rubric, with blind subagent validation for self-scoring bias, plus a `freshen` mode that probes external references (release notes, docs, deprecation signals) and applies verified updates, plus a `trigger` mode that measures and tunes the skill's frontmatter description until it reliably fires when it should and stays silent when it shouldn't (60/40 train/test split, 3 runs/query, blinded test scores).

Create and track a DeepVista workflow skill for the current Claude Code session. On first invocation, captures the session goal as a note and synthesizes a workflow skill via deepvista skill create-from-note. Subsequent invocations in the same session reuse the existing skill. A Stop hook auto-syncs node status, timestamps, error details, and session metrics to the skill card after every turn.

Write, review, or refactor Rust so it is idiomatic, safe, and maintainable. Use this skill whenever the task involves Rust code — implementing a feature in a `.rs` file, reviewing a Rust PR or module, refactoring a crate, designing a public API, fixing clippy or borrow-checker complaints, deciding how to structure error handling, or splitting an oversized function or file. Trigger even when the user does not say "best practices" explicitly: any request to "write this in Rust", "clean up this Rust code", "is this idiomatic", "add a Cargo crate", "handle these errors properly", or "make this compile without unwrap" should pull in this skill. Covers ownership and borrowing, error handling (Result / `?` / thiserror / anyhow), API design, module and crate layout, size limits, unsafe, dependencies, testing, and the verification steps (fmt, clippy, test) that must pass before the work is called done.