arendon1

@arendon1 source repo

44 skills
0 followers
6 hours ago last updated

▌
Git · arendon1
Proactive version control. Creates worktrees for plan work, commits at checkpoints with conventional commits, merges at plan completion, resolves conflicts safely. Use when starting a session, hitting a plan checkpoint, or finishing a branch, or when the user says "commit", "checkpoint", "merge", "branch", "worktree", "resolve conflict".

0
installs
▌
Tdd · arendon1
Test-driven development. Write the test first, watch it fail, write minimal code to pass, refactor. Vertical slices via tracer bullets — one test to one implementation, never all-tests-then-all-code. Tests verify behavior through public interfaces, not implementation. Use when implementing any feature or bugfix, before writing implementation code, or when the user says "implement", "build this", "add a feature".

0
installs
▌
Plan · arendon1
Break a PRD/ARD/SPEC into an implementation plan of vertical-slice tasks. Each task cites the invariants and interfaces it touches, names the test that proves it, and has a status cell. Sole mutator of PLAN.md in the active plan. Use when ready to break requirements into tasks, after grill/design/spec and before build, or when the user says "plan this", "break this down", "tasks", "what's the plan".

0
installs
▌
Spec · arendon1
Detailed technical specification. Turn a PRD/ARD into a precise SPEC.md: exact schemas, algorithms, data shapes, edge cases, acceptance criteria. Sole mutator of SPEC.md in the active plan. Has NEW / AMEND / DISTILL modes. Use when the objective needs more detail than the ARD before planning, after design, or when the user says "spec this", "detailed spec", "specify the behavior".

0
installs
▌
Check · arendon1
Read-only drift detector. Diffs the plan artifacts against the current code and reports violations grouped by severity. Writes nothing — suggests remedies but never invokes them. Use when checking drift, auditing the spec, verifying invariants hold, or when the user says "check drift", "audit the spec", "does the code still match", "spec vs code", "is the plan still accurate".

0
installs
▌
Debug · arendon1 bundle
Disciplined diagnosis loop for hard bugs and performance regressions. Build a tight feedback loop first, reproduce, minimize, hypothesize, instrument, fix, regression-test. Never guess at a fix before root cause is found. Use when the user says "diagnose" or "debug this", or reports something broken, throwing, failing, or slow, or when a build verification fails.

0
installs
▌
Grill · arendon1
Sharpen a fuzzy idea into requirements before spec. Calibrated interrogation via the grilling discipline: frontier rounds, recommended answers, facts are the agent's job, land each decision in PRD.md or CONTEXT.md. Creates the plan folder. The cheapest place to kill a bad idea is before a task exists. Use when the user has a vague idea, says "grill me", "stress-test this", "challenge my plan", "interview me", or before designing anything non-trivial.

0
installs
▌
Teach · arendon1
Teach the user a new skill or concept over multiple sessions, using the current directory as a stateful teaching workspace. Build exercises, check understanding, adapt pace to the learner. Use when the user wants to learn something, not build something, or says "teach me", "I want to learn", "explain over time", "help me understand", "course on", "tutorial".

0
installs
▌
Use Agy · arendon1 bundle
Use the Antigravity CLI (`agy`) as a worker or third-opinion reviewer from any agent harness. `agy` is Google's agentic coding-agent backend (same family as `codex` / `claude-code`); auth is OS-keyring + browser Google sign-in, so its **Gemini models ride on Google AI Pro at zero marginal token cost**. Reach for `agy` for a cheap Gemini answer, a Claude-family second opinion on another agent's plan or diff, or a sandboxed multi-turn coding session outside the host's model menu. `agy` is a worker, not an orchestrator — route work through the normal task graph. Use when running a one-shot `agy -p` review or change, launching an interactive `agy -i` session, fanning a parallel batch across git worktrees with one `agy -p` each, picking a Gemini model on a Google AI Pro budget, asking for a Claude Sonnet/Opus second opinion on a plan or diff, inspecting `~/.gemini/antigravity-cli/`, or debugging auth/sandbox/permission state.

0
installs
▌
Deepen · arendon1
Make one shallow module deeper — move complexity behind a smaller interface, held at a clean seam, testable through that interface. Behavior preserved, tests green before and after. Proposes §I/§V/§T edits, never silent rewrites. Use when a module is shallow (big surface, little behavior), or when spare budget exists for a design pass, or when the user says "refactor this", "this is messy", "deepen", "improve this module".

0
installs
▌
Design · arendon1
Architecture design pass. Turn a PRD into an ARD: components, interfaces, invariants, data flow, error handling. Propose 2-3 structural approaches with trade-offs. Recommend one. Behavior preserved. Use when the objective is non-trivial or high-blast-radius, after grill and before spec, or when the user says "design this", "architecture", "how should we structure this".

0
installs
▌
Review · arendon1
Adversarial senior review of the plan before build, or of the code after build. Constructs a skeptical reviewer anchored to the codebase, RESEARCH.md, and live best-practice, then tries to REFUTE — not rubber-stamp. Every finding cites evidence. Code review runs two axes (Standards + Spec) as parallel isolated reviewers, reported side by side. Ends in a go/no-go gate. Use when reviewing high-blast-radius work before build, or reviewing the diff after build, or when the user says "review the spec", "red-team this", "is this sound", "senior review".

0
installs
▌
Triage · arendon1
Move issues through a triage state machine: from raw inbox to labeled, scoped, and ready for planning. Define triage roles/labels, apply them consistently, and surface what is ready to become a plan. Use when managing a backlog of issues or tickets before they become plans, or when the user says "triage", "go through issues", "organize the backlog", "what should I work on", "prioritize issues".

0
installs
▌
Verify · arendon1
Completion gate. Ensure work is actually done before declaring it complete. Verify against the plan, run the full suite, confirm no regressions, check artifacts are updated. Evidence before assertions, always. Use when about to mark a task, plan, or objective complete, or when the user says "done", "complete", "finished", "ready to ship", "is this done".

0
installs
▌
Use Cmux · arendon1 bundle
Drive the cmux terminal multiplexer from an agent session: give each long-lived process (dev server, build watcher, test runner) its own pane, open localhost webapps in browser surfaces, run parallel work in dedicated surfaces, launch peer agent sessions with isolated cwd/output/lifecycle, send commands, read output, and signal completion via notifications. Use when the agent must run a long-lived process in a dedicated pane, preview a localhost URL in a browser panel, give a parallel task or peer agent its own surface, monitor or control a pane's output, manage terminal topology across windows/workspaces/panes/surfaces, or coordinate multiple agent sessions that each need their own terminal.

0
installs
▌
Use Hunk · arendon1 bundle
Orchestrate a pause-for-review workflow around the hunk terminal diff viewer. Decides when to pause (keyword + risk-aware), opens hunk in the active terminal multiplexer (cmux, herdr, or instructs the operator to launch it manually), prints a chat preamble, leaves selective inline notes on non-obvious decisions, waits for the operator's next message, and classifies it as approve / feedback / stop / ambiguous before proceeding. Use when the agent has just written or modified code and the operator may want to review it, when a change crosses the risk threshold (>= 80 lines, >= 4 files, auth/schema/migration/secret/lockfile/ci paths, new dependency, revert), when the operator says "revisar|review|pausa|para|hold|wait", when the agent is about to commit or push, when the operator asks to walk them through a changeset, or when the agent is not sure whether the operator wanted a review or autonomy and should flag the ambiguity.

0
installs
▌
Handoff · arendon1
Compact the current session into a handoff document so another agent (or a fresh session) can continue the work. Captures goal, state, decisions, next steps, and pointers to artifacts. Does not duplicate content already in artifacts. Use when ending a session mid-plan, or when handing work to another agent, or when the user says "handoff", "compact this", "I'm leaving", "pass this to another agent", "save state for later", "wrap up".

0
installs
▌
Lessons · arendon1
Bug to spec protocol. When a bug is found or a test fails during build, trace the root cause, decide whether a new invariant would catch recurrence, and append a lesson to the active plan's LESSONS.md. This is the reflex that stops classes of bugs from recurring. Use when a test fails, a bug is reported, a post-mortem is run, or a build verification fails.

0
installs
▌
Caveman · arendon1
Compressed English encoding for plan artifacts and spec-adjacent prose. Cuts tokens ~75% vs prose while staying readable. Applies to PRD.md, ARD.md, SPEC.md, PLAN.md, LESSONS.md, RESEARCH.md, CONTEXT.md. Use when writing to a plan artifact, spec-adjacent prose, or when the user says "caveman", "compress this", "be brief", "write to PLAN.md".

0
installs
▌
Use Herdr · arendon1 bundle
Drive the herdr agent multiplexer from an agent session: give each long-lived process (dev server, build watcher, test runner) its own pane, run parallel work in dedicated panes/tabs/workspaces, launch peer agent sessions with isolated cwd/output/lifecycle, send commands, read output, wait on agent state changes, and signal completion via notifications. Cross-platform (Linux, macOS, Windows) — the portable alternative to cmux when you are not on macOS or not inside a cmux terminal. Use when the agent must run a long-lived process in a dedicated pane, preview a localhost URL (by opening a browser pane in the terminal), give a parallel task or peer agent its own pane, monitor or control a pane's output, wait on another agent's state, manage terminal topology across sessions/workspaces/tabs/panes, coordinate multiple agent sessions that each need their own terminal, or check out a git worktree as a workspace for parallel branch work.

0
installs
▌
Dispatch · arendon1
Dispatch fresh subagents for independent tasks, with two-stage review (spec compliance, then code quality). Use when a plan has independent tasks that can run in parallel, or when a task is large enough to benefit from an isolated subagent. Use when the user says "parallel", "dispatch", "subagent", or when the plan has independent tasks, or when a task is large enough to benefit from isolation.

0
installs
▌
Grilling · arendon1
Relentless interview primitive: map the design tree, work the frontier in rounds — every question whose prerequisites are settled gets asked in one round, numbered, each with a recommended answer. Facts are the agent's job, never the user's. Use when any loop needs to sharpen a fuzzy idea by interview (grill, triage, deepen, wayfinder), or when the user asks to be grilled, stress-tested, or challenged.

0
installs
▌
Research · arendon1
Gather external knowledge into the active plan's RESEARCH.md so build grounds in facts, not hallucinations. Every finding cites a source. Delegates to domain capabilities (e.g. research-literature for academic) based on their provides field. Use when the plan needs external facts before building, or when the user says "research this", "look up", "find docs on", "is there a library for".

0
installs
▌
Get Docs · arendon1
Ground the active agent in the project's tech stack by fetching live documentation via the Context7 CLI (`ctx7`, REST API fallback) and materializing relevant snippets as markdown references under `docs/stack/<lib>/<topic>.md` in the active project. Run after the stack has stabilized (typically after `design` or `spec` chose a library) so the same docs don't get re-fetched on every session. Use when the user says "ground the stack", "materialize docs for X", "fetch docs for this project", or asks about a library the active project uses without a `docs/stack/<lib>/` folder yet.

0
installs
▌
Make A Map · arendon1 bundle
Skill de información geográfica y renderizado SVG vectorial. Resuelve consultas de ubicación, descarga geometría administrativa + calles + parques + cuerpos de agua + edificios vía OSM (Nominatim + Overpass), aplica fallback automático a Mapbox cuando OSM falla o devuelve datos insuficientes, proyecta a Web Mercator (EPSG:3857) y emite un SVG limpio agrupado en <g id="..."> semánticos listos paraAffinity Designer, PowerPoint o Illustrator. Usa cuando el usuario pida un mapa vectorial editable de una ciudad/barrio/región, un plano de calles, un mapa ilustrativo para una presentación o material académico, o cuando entregue coordenadas explícitas y necesite un SVG geográficamente fiel. Use when the user asks for a vector map, SVG export of a city, editable street map, or geographic visualization in SVG form.

0
installs
▌
Agents Pm · arendon1 bundle
Agent project management over the shared ClickUp **Agents** space: keep every agent's work visible, scoped, claimed, and non-overlapping across devices and harnesses. Statuses drive progress; lists enclose big named projects or goals. Use when starting a session or after context compaction, before picking up any non-trivial work, to check what other agents are doing, to register new work before it starts (nothing is built outside scope), to claim, update, block, review, or close a tracked task, or when asked about the state of the team's work.

0
installs
▌
Prototype · arendon1
Build a throwaway prototype to flesh out a design — a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route. Cheap exploration before committing to a spec. Use when a design question is cheaper answered by building than by talking, or when the user says "prototype", "spike", "let's try it", "build a throwaway", "explore this design", "proof of concept", "POC".

0
installs
▌
Wait What · arendon1
Corrective for a message that did not land. When the user says "wait what", "I don't understand", "that makes no sense", "re-pitch", or otherwise signals confusion mid-conversation, stop and re-pitch the last point with the missing context, in plain language, using the CONTEXT.md vocabulary. Use when a message missed, the user seems lost, or the user asks to re-pitch, rephrase, or explain again.

0
installs
▌
Bootstrap · arendon1
Session enabler. Teaches the agent to check for and invoke relevant skills before any response, and to check for project artifacts (CONTEXT.md, active plan folder). Use when a session starts or after a compaction resets context. Injected at session start and after compaction. The only bootstrap-invoked skill.

0
installs
▌
Skill Add · arendon1
Integrate an external skill into this repo. Fetch via npx skills add, audit against the constitution with skill-forge, place with upstream license attribution, and report incompatibilities. Use when adding an external skill, after skill-find picks one, or when the user names a source to add. Use when the user says "add this skill", "install skill", "integrate skill", "pull in skill", "skill-add", or names a source to add.

0
installs
▌
Use Clickup · arendon1 bundle
ClickUp task and list management via the official API. Use when creating or updating ClickUp tasks (due dates, statuses, assignees), searching task activities and history, creating lists, or viewing existing lists. Handles authentication, response caching, and known API quirks so agent-driven ClickUp edits land correctly.

0
installs
▌
Check Subs · arendon1 bundle
Probes 4 AI provider subscriptions (OpenRouter, opencode-go Go, MiniMax, Google AI Pro) and reports liveness and remaining quota per usage window. Writes a state file that any downstream consumer can read to make routing decisions. Use when an agent is about to dispatch a task to a subscription-based model, after a 429/2056/quota-exhausted response, when the user asks for current subscription state, or before opening a long working session.

0
installs
▌
Skill Find · arendon1
Discover external skills from the skills.sh registry. Search by keyword and/or GitHub owner, present ranked options with install counts, and recommend. Does NOT install — pair with skill-add. Use when the user wants to find a skill for a task, or says "find a skill", "is there a skill for", "search skills", "browse skills", "what skills exist for".

0
installs
▌
Skill Forge · arendon1 bundle
Constitution enforcer for the agent-skills repo: scaffolds new skills with compliant frontmatter (init), validates any skill against AGENTS.md (audit — frontmatter schema, layers, triggering contracts, agnosticism, size), and regenerates or verifies the harness-discovery manifest (manifest). Use when creating a new skill, auditing a skill for compliance, migrating a skill to the new frontmatter schema, or checking that the manifest is fresh.

0
installs
▌
Analyze Model · arendon1 bundle
Fetches live model data from OpenRouter and Artificial Analysis APIs, merges benchmarks with pricing, runs deep cost analysis on usage logs, exports usage from LLM surfaces (OpenCode, Pi, VS Code, etc.), analyzes subscription efficiency, and forecasts future spend with cheaper-model alternatives. Use when comparing LLM costs or quality, analyzing total API spend from usage logs, exporting agent usage for cost analysis, evaluating subscription cost efficiency, forecasting future LLM costs, finding cheaper alternatives, or when the user asks about model pricing, intelligence index, speed, or token costs across providers.

0
installs
▌
Generar Paper · arendon1 bundle
Genera documentos académicos en formato APA 7 (nivel estudiante) con validación estricta de citas y referencias, y salida en PDF vía Typst. Usa cuando el usuario necesite validar que las citas estén correctas y completas, formatear bibliografía o referencias en APA 7, componer o estructurar un paper (IMRaD: introducción, métodos, resultados, discusión), o compilar un documento académico a PDF final.

0
installs
▌
Use Firecrawl · arendon1 bundle
Spin a fully local containerized Firecrawl instance wired to OpenRouter for LLM + embeddings, verify Docker can mount it, and wire it to whatever harness the agent is running in (Hermes, Pi, OpenCode, or generic). Use when the user says "spin a local Firecrawl", "self-host Firecrawl", "give me web crawling", "set up web_search/web_extract locally", or wants crawling/scraping without paying Firecrawl SaaS. Defaults: deepseek/deepseek-v4-pro for LLM, qwen/qwen3-embedding-8b for embeddings — both via OpenRouter. Asks for the OpenRouter API key interactively. Detects the harness and writes the wiring to the right place.

0
installs
▌
Make A Diagram · arendon1 bundle
Generate software architecture diagrams, flowcharts, sequence diagrams, block diagrams, C4 models, ER schemas, Sankey flows, state machines, GitGraphs, timelines, and more using Mermaid (v11+) with ELK layout engine. Renders SVG, PNG, and ASCII art via beautiful-mermaid. Use when users ask to diagram, visualize, or map software concepts, system architecture, data flows, processes, or database schemas.

0
installs
▌
Gestionar Cursos · arendon1 bundle
Extrae y sincroniza cursos Moodle Uniremington (Aula Virtual): inicializa desde URLs, organiza la estructura local y verifica estado vs Moodle. Reutiliza la sesión persistente en `.browserdata/` — NO pidas re-login si hay sesión válida. Usa cuando necesites iniciar un curso (init), sincronizar contenido o calificaciones con ClickUp, revisar foros evaluables, consultar calificaciones, verificar cambios (estado), o el usuario diga que ya inició sesión.

0
installs
▌
Use Mineru Cloud · arendon1 bundle
Convert PDFs, images, DOCX, PPTX, and XLSX documents into clean Markdown via the MinerU cloud API (mineru.net) — headers/footers stripped, reading-order text, tables→HTML, formulas→LaTeX, OCR for scanned pages, images extracted alongside. Use when the user hands you a PDF (or document/image) to read, summarize, extract, or feed into a knowledge base; when they say "convert this PDF to markdown", "parse this document", "extract this paper", "make this readable"; when OCR of a scanned PDF is needed; when running on a lightweight device (phone/tablet) where local heavy parsing is impossible; or when batch-converting many documents at once. Two modes: flash (no token, fast, ≤10MB / ≤20 pages) and precision (Bearer token, full assets, ≤200MB / ≤200 pages, local-file upload). Never stores secrets — the token is read from the MINERU_TOKEN environment variable only.

0
installs
▌
Domain Modeling · arendon1
Actively build and sharpen a project's ubiquitous language. Challenge terms against the glossary, stress-test with edge-case scenarios, update CONTEXT.md inline the moment a term crystallizes, offer ADRs for hard decisions. Use when terminology is fuzzy, overloaded, or conflicts with the glossary, or when the user says "what do we call this", "define this", or asks about ubiquitous language.

0
installs
▌
To Questionnaire · arendon1 bundle
Turn a decision the user cannot answer alone into a questionnaire for someone else to fill in. The inverse of grilling: interview the user about the SEND (who it goes to, what they need back), then write questions that target the gap between what the recipient knows and what the user needs. Use when the blocker is knowledge in another person's head — a stakeholder, client, expert, or colleague — and the user needs them to answer async or in a meeting.

0
installs
▌
Compensation Report · arendon1
Entrevista al usuario para recolectar los datos del cargo objetivo (cargo(s), años de experiencia, ubicación/modalidad, empresa objetivo y stack tecnológico), investiga bandas salariales en vivo en múltiples fuentes, triangula los datos y genera un reporte de compensación en markdown con valores en COP y USD, anuales y mensuales, perfil financiero del empleador, estructura de compensación esperada, BATNA implícito y tres tácticas defensivas de negociación. Usa cuando el usuario pida un reporte de compensación, un análisis salarial para una negociación, bandas salariales para un cargo en una ubicación específica, o quiera preparar una conversación salarial con un empleador actual o potencial.

0
installs
▌
Research Literature · arendon1 bundle
Busca artículos académicos y fuentes científicas usando EXA, Semantic Scholar y Google Scholar, con resultados estructurados para citar en papers. Usa cuando el usuario necesite encontrar fuentes para un paper o ensayo, revisar el estado del arte de un tema, preparar una revisión de literatura, o diga "busca papers sobre...", "encuentra artículos de...", "qué investigaciones hay sobre...", "fuentes para mi paper".

0
installs

44 published skills