agents-skills

USER-INVOKED ONLY. Do NOT auto-trigger. Convene 3 fresh Codex subagents to give independent opinions on a question, idea, hypothesis, plan, or ambiguity. Use only when the user explicitly invokes this skill ("$council", "/council") or explicitly names it ("run council", "use the council skill", "convoque le council"). Never trigger on musings, opinions, or phrases like "je me demande si" — thinking out loud is not an invocation. If invoked with no extra prompt, infer the current subject from the active conversation and ask clean-context subagents to review it.

Drive and visually inspect a native GUI app (Iced, egui, GTK, winit…) on Hyprland through a headless output, leaving the user's desktop untouched (focus preserved, cursor restored). Use when asked to "test the GUI", "piloter l'app native", "screenshot the app", "voir le rendu", to verify a native app's rendering after a change, or to interact (keys, text, clicks) with a native window that exposes no AT-SPI tree. Do not use for web or Electron apps (use agent-browser), apps exposing a usable AT-SPI accessibility tree (prefer a semantic tool), or compositors other than Hyprland.

Use when building, operating, or resuming a deterministic Workflow() driver for a large autonomous multi-phase task — a full rewrite, a migration, a long audit. Triggers: design or implement a workflow driver script, translate a .agents/loops/*.md runner spec (looper output) into a Workflow() script, design phase gates or pause/resume mechanics, handle a paused driver run (retryStep bump, human arbitration, mutation suspected), harden a driver against non-deterministic agent behavior, or wrap an external CLI (e.g. codex exec) inside a workflow agent.

USER-INVOKED ONLY. Use only when Augustin explicitly invokes $skill-sync or asks to commit, push, and deploy agent skill changes live from agents-skills. Commits scoped source, pushes the default branch, publishes exact managed skill names through the shared post-push gate, and verifies live parity without touching project-pinned copies.

Portable discipline for designing, running, and interpreting skill evaluations with skill-creator. Use when defining eval success criteria, running or comparing trials, grading outputs or end states, reviewing traces, benchmarking a skill, iterating on evals, or maintaining eval artifacts. Separates experimental validity from artifact integrity and harness-specific formats.

Write, review, and harden non-trivial Bash and POSIX sh scripts for production. Use for shell files and CI, deployment, administration, or automation code involving untrusted input, destructive filesystem operations, privileges, pipelines, traps, signals, concurrency, portability, or failure handling. Skip trivial interactive commands and tasks better served by an existing project tool.

Generate a handoff prompt for another agent instance. Use when the user asks to write a prompt, meta-prompt, or briefing for a separate agent session — phrases like "write a prompt for another instance", "make a handoff prompt", "brief another agent", "prepare a prompt I can copy-paste", "write this up for another session", or when the user has been discussing/planning a task and wants to delegate execution to a fresh agent instance. Also triggers on "handoff", "meta-prompt", "delegate to another agent".

USER-INVOKED ONLY. Use only when the user explicitly invokes $premium-handoff or asks to turn a selected target, targetify seed, evidence summary, design/architecture problem, or premium-model target into a compact handoff prompt/artifact for a scarce frontier model. Do not target the whole repo, implement, or launch the premium model.

USER-INVOKED ONLY. Use only when the user explicitly invokes $targetify or asks where to spend scarce/premium/frontier/super-model attention in a repo. Read-only. Do not use for implementation, full audits, premium handoffs, or launching premium model runs.

Human-invoked meta-skill for turning a broad intention into staged local artifacts: intake, clarification, terminology, PRD, issues, workflow, tracker, prompts, and final execution handoff. Use only when the user explicitly invokes $intent-to-workflow or asks to scaffold this PRD-to-issues-to-workflow process with human gates.

USER-INVOKED ONLY. Use when the user explicitly invokes $looper or /looper or asks to design, audit, review, or convert a task into a Codex loop: design a loop, make this a loop, turn this into a Codex loop, Theo loop, Pete loop, loopify, inner loop, watch this PR, loop until review passes, or create a dynamic workflow. Do not invoke for normal implementation, review, or planning unless the user asks for a loop.

USER-INVOKED ONLY. Invoke only when the user runs $second-brain or explicitly names this skill. Stand up and maintain an agent-maintained markdown wiki that loads progressively and stays bounded: a thin always-loaded index plus a recent-window journal, with the source catalog, sealed period archives, and synthesis pages pulled only on demand. Triggers: the user invokes $second-brain; wants to set up a second brain, knowledge base, agent memory, or LLM wiki; or a memory/log/notes file has grown large and is reloaded in full every session and they want it windowed and bounded. Not for a handful of notes, and not for a million-document corpus (that is a search/RAG problem). Never invoke implicitly.

Convert Mermaid code blocks in Markdown files to PNG images using the mermaid.ink API.

USER-INVOKED ONLY. Use when the user explicitly invokes $slice-runner or asks Codex to execute an already-approved multi-slice implementation plan with native Codex subagents. Requires runnable quality gates. Do not use for planning; use intent-to-workflow first when the plan is not yet agreed.

Detect, diagnose, score, and rewrite AI-generated or generic text. Trigger on: "check this for slop", "does this sound AI", "rewrite this", "make this sound human", "edit my LinkedIn post", "fix my cover letter", "this sounds generic", "clean up my writing", "audit this draft", "too much fluff", "sounds like ChatGPT", "deslop", "voice check", or any request to improve, critique, or humanize written text. Works in English, French, and Spanish — also trigger on French/Spanish cues like "ça fait IA", "détecte le slop", "rends ça plus humain", "trop d'anglicismes", "suena a IA", "esto suena a ChatGPT". Also trigger when reviewing any text output from another agent or skill that will be published, sent, or shared externally.

Teach a human through the current session incrementally until they can demonstrate real understanding. Use when the user invokes "$teach-back", asks to be taught or debriefed on a session, wants a running understanding checklist, asks for teach-back verification, ELI5/ELI14/ELII explanations, quizzes, or a mastery-gated walkthrough of problem, solution, edge cases, and broader impact.

Use this skill when the user wants to convert rough coding, product, architecture, audit, review, migration, cleanup, debugging, or orchestration intent into a compact Codex `/goal` payload. Trigger on "goalify", "Codex goal", "/goal", "make this a goal", "RALF loop", "turn this into a Codex prompt", "$goalify interactive", or requests to be questioned before creating a goal. Do not use for normal implementation or review unless the user explicitly asks to produce a Codex goal/prompt.

USER-INVOKED ONLY. Use only when Augustin explicitly invokes $promptify or explicitly names this skill with one raw prompt argument. Transform a rough dictated instruction into a clean GPT-5.5-optimized prompt. If blocking ambiguities remain, do not produce the final prompt; route to $grill-me first.

USER-INVOKED ONLY. Use only when the user explicitly invokes $of or $op, or explicitly names this skill to open/view a local file. Never invoke implicitly. Opens visualizable files such as images, Mermaid, Markdown, HTML, PDF, and text using the system default app.

Run a post-feature polishing loop over the current WIP or branch diff. Use when the user wants Codex to simplify recently generated code, reduce slop, make TypeScript code more expressive with justified types/helpers, or run clean-context reviewers before accepting a change set.

Use when read-only research clearly splits into multiple independent code, docs, logs, tests, or web questions worth parallel subagents and concise synthesis.

USER-INVOKED ONLY. Use only when the user explicitly invokes $docs-drift or asks to audit Codex guidance drift. Audits AGENTS.md, AGENTS.override.md, .agents/skills, .codex/config.toml, project skill guidance, and other Codex instructions against recent code changes. Inspect real files first; use official OpenAI Codex docs or live web only for current Codex behavior; report findings before edits.

USER-INVOKED ONLY. Use only when the user explicitly invokes $skills-maintenance or explicitly asks to update installed agent skills globally, update project skills, refresh skills-lock files, reconcile dotfiles-managed skills.sh/bootstrap state, or commit/push skill-update maintenance changes. Runs skills.sh updates through a bundled script, then verifies source/live/dotfiles boundaries before any commit.

Claude Code hooks authoring and audit skill. Use when creating, auditing, modifying, or explaining Claude Code hooks, .claude/settings.json hook configuration, hook events, handler types, matchers, testing, debugging, or official hooks examples. Focuses on current official docs plus small local scripts; not for generic Git hooks.

Use when designing explicit state-machine behavior for systems with multiple modes (idle/loading/success/error, draft/submitted/approved, connected/disconnected/reconnecting), boolean-flag explosion where isLoading, isError and isSuccess drift out of sync, forms, wizards, auth flows, network protocols, embedded controllers, workflow engines, or any code/bug report where "impossible states" appear. Agent-facing design workflow: model states, events, transitions, guards, actions, and impossible states before implementation. Provides implementation references for TypeScript discriminated unions, XState v5, React useReducer, Svelte 5 runes, C, Java, and Rust. NOT for installing tools, single-boolean toggles, pure server-cache concerns better handled by TanStack Query or SWR, or stateless pure transformations.

USER-INVOKED ONLY. Use only when the user explicitly invokes $pr-review-html or explicitly names this skill. Create a self-contained HTML artifact for reviewing a pull request, patch, or local diff with real diff excerpts, inline margin annotations, severity-colored findings, reviewer handoff context, and explanations for unfamiliar code paths such as streaming, backpressure, concurrency, retry, cancellation, or state-machine logic. Never invoke implicitly.

TypeScript CLI scripts and applications using the Bun runtime. Covers file structure conventions, Bun Shell ($), File I/O, testing with bun:test, HTTP servers, SQLite, and bundling. NOT for: Node.js/Deno projects, frontend-only React/Vue apps, or npm package publishing workflows.

USER-INVOKED ONLY. Use only when the user explicitly invokes $batch-update or explicitly names this skill to update a repository. Create an isolated git-wt worktree, inventory repo update surfaces, research and apply dependency/tooling update batches with native ecosystem tools, validate, commit each successful batch, and document deferrals. Never invoke implicitly.

Create a complete, actionable design system for any project (website, app, product) grounded in physical/sensory anchoring rather than design trends. Delivers CSS tokens, typography scale, component patterns, accessibility checklist, and absolute rules as a single developer-ready markdown document. Use when the requested deliverable is a design system, style guide, visual identity system, CSS/theme tokens, color and typography system, component styling rules, or structured art direction. Do not use for one-off UI styling, generic look-and-feel feedback, page implementation, content architecture, PRDs, or implementation tickets.

Design content architecture for digital products when the needed output is a sitemap, page map, navigation structure, screen inventory, content hierarchy, or CLI-to-frontend mapping. Use when deciding what pages/screens exist and how they connect. Do not use for UI implementation, visual design, design systems, complete PRDs, implementation tickets, or code architecture.

Human-invoked architecture workflow for finding and designing deeper codebase modules. Use when the user wants an architecture audit, interface design, or RFC issue focused on locality/colocation, leverage, clear seams, and reducing AI navigation entropy. Not for broad cleanup, style refactors, or immediate implementation.

Interactively shrink AI agent skills while preserving their useful semantics. Use when the user wants to reduce, compact, slim, shrinkify, deslop, or optimize a SKILL.md or skill folder, especially when they want fewer tokens without losing triggers, output shape, decision rules, validation behavior, or recurring workflow quality. Supports explicit audit and apply modes.

Tooling automation for @effect/language-service in Effect projects. Runs CLI commands (diagnostics, quickfixes, codegen, overview, layerinfo), interprets diagnostic output, applies automated fixes, configures rule severities, and sets up CI integration. Use when working with Effect diagnostics, fixing ELS warnings, generating code from directives, analyzing layer composition, or configuring tsconfig plugin options. NOT for: architectural decisions about Effect adoption (use effect-usage-cc), non-Effect TypeScript, or manual code review without ELS tooling.

Decision support for Effect (effect-ts) in TypeScript projects. Determines when to use Effect, which patterns to prefer, and where plain TypeScript is the better choice. Invoke when reviewing architecture, writing Effect code, evaluating adoption, or questioning whether Effect fits a task. NOT for: plain TypeScript without Effect, React component logic, simple Bun scripts with no error recovery.

Generate pixel-perfect ASCII box diagrams for architecture docs, READMEs, and code comments. Use when the user asks to create, fix, or improve ASCII diagrams, draw architecture diagrams, create box-and-arrow diagrams for documentation, or when you need to add diagrams to markdown files. Also use proactively when generating architecture documentation that would benefit from visual diagrams. Triggers on: "draw a diagram", "ASCII diagram", "box diagram", "architecture diagram", "add a diagram to the doc", "fix the alignment", "the boxes are misaligned", or any request to visualize system architecture in text.