JairoTorregrosa

@jairotorregrosa source repo

12 skills
0 followers
21 hours ago last updated

▌
Askcodex · jairotorregrosa bundle
Use OpenAI's GPT-5.x and image models from the command line with the `askcodex` binary: ask or complete text, generate or edit images, list models, check quota/plan. Trigger whenever the user wants an OpenAI model from the terminal — "pregúntale a gpt-5", "haz una imagen", "qué modelos tengo", "cuánta cuota me queda", "quién soy" — or when a quick one-shot task (summarize, classify, draft, generate an asset) is worth delegating to an OpenAI model. Auth is already on the machine; there is nothing to configure. Read this file completely before running anything: the flags, output locations and gotchas are below, and references/prompting.md carries the per-model, per-use-case prompting guide.

0
installs
▌
File Todos · jairotorregrosa bundle
File-based TODO lifecycle for tracking review findings, managing technical debt, and organizing work items. Each TODO is a standalone markdown file with YAML frontmatter, named by convention: {id}-{status}-{priority}-{description}.md. Status transitions drive the lifecycle: pending (created from review) -> ready (triaged and approved) -> complete (resolved). Use when: (1) review findings need persistent tracking beyond a single session, (2) technical debt should be cataloged with solutions, (3) work items need triage before implementation. Do NOT use for: (1) simple one-off fixes that can be done immediately, (2) tasks already tracked in an external issue tracker.

0
installs
▌
Metaprompt · jairotorregrosa bundle
Generate a complete, ready-to-use prompt for a target model and harness. Triggers: metaprompt, generate a prompt for, write me a prompt, create a system prompt, prompt engineer this, optimize this prompt

0
installs
▌
Insistir · jairotorregrosa bundle
Agents Checking Agents — multi-agent orchestration with cross-validation using Claude Code Agent Teams and a cross-provider Codex/GPT-5 judge for bias-free final verdicts. Fresh agents for each phase — no agent reuse. Each worker/reviewer/fixer is spawned, does its work, and is shut down. The review loop iterates until the reviewer APPROVES or max rounds are reached. The lead operates in delegate mode: coordinates only, never implements. Use when: (1) user says /insistir, (2) complex tasks that benefit from parallel execution with quality gates, (3) "agents checking agents", "multi-agent", "swarm", or "orchestrate". Do NOT use for: (1) simple single-file changes, (2) sequential tasks with tight dependencies, (3) quick bug fixes or typos, (4) tasks where coordination overhead exceeds the benefit. Full pipeline: intake → plan → spawn team → [execute wave → adversarial review → codex judge] → cleanup.

0
installs
▌
Agent Sdk Wizard · jairotorregrosa bundle
Build a working Claude Agent SDK agent by asking the user 9 illustrated questions, one at a time, then generating and running the code. Trigger on "crea un agente", "quiero hacer un agente con el SDK", "ayúdame a armar mi agente", "nuevo agente paso a paso", "agent sdk wizard", "build an agent with the Claude Agent SDK", "help me create an agent step by step", "set up an agent project". Each question is one AskUserQuestion call whose options all carry an ASCII diagram in `preview`, in plain language for someone who has never seen an agent. Produces a directory with agente.py or agente.ts, its manifest, a README, a decisiones.md, and a real test run. NOT for the raw Messages API, Managed Agents, or the Claude Code CLI itself — those are different products; say so and stop.

0
installs
▌
Goal Loop · jairotorregrosa bundle
Loop engineering as gradient descent — a guided goal loop where an agent factory generates goal-specialized agents (implementer, verifier, diagnoser, judge) and iterates forward → loss → backward → update until the goal is provably met. Evidence is split into visible validation (the implementer's loss) and held-out checks (judge-only, anti-reward-hacking). A diagnoser turns failures into textual gradients; momentum accumulates recurring patterns; plateau detection triggers early stopping. A Phase 0.5 observability plan reuses or generates the instruments needed to verify the goal, calibrated red-first before the loop starts. Use when: (1) user says /jaiskills:goal, (2) "loop engineering", "goal loop", "agent factory", "gradient descent on a task", "loop until done", "iterate until passing", (3) a task needs verifiable completion evidence across multiple attempts. Do NOT use for: unverifiable goals (refine with the user first), trivial one-shot tasks, or open-ended exploration with no done-state.

0
installs
▌
Image To Frontend · jairotorregrosa bundle
This skill should be used when the user asks to "image to frontend", "build this UI", "make me a landing page", "design and ship this site", "generate a mockup", "turn this screenshot into code", or hands over a product brief or reference image and wants a working React or HTML page. Also covers single-shot app screenshots (desktop + mobile pairs), logo and hero illustration generation, photorealistic compositing, and identity-preserving image edits via the askcodex CLI gpt-image path (no API key required; uses `codex login` credentials).

0
installs
▌
Codex Judge · jairotorregrosa bundle
Cross-provider LLM judge using Codex (GPT-5) to evaluate implementations written by Claude. Internal skill invoked by the insistir lead during the review loop. Not user-facing.

0
installs
▌
Remote Agents · jairotorregrosa bundle
Orchestrate headless Claude Code and Codex agents on a remote SSH host (Jetson) from this Mac. Use when: (1) user says /jaiskills:remoto-run or /jaiskills:remoto-status, (2) 'remote agents', 'spawn on the jetson', 'run this on the jetson', 'remote swarm', (3) work should execute on the remote box (its GPU, its files, its environment) while this session coordinates. Do NOT use for local-only tasks or when the remote host is unreachable.

0
installs
▌
Compound Knowledge · jairotorregrosa bundle
Document a solved problem to compound team knowledge. Triggered by phrases like "that worked", "it's fixed", "problem solved", or explicitly via the /jaiskills:compound command. Orchestrates parallel sub-agents to analyze the problem, extract the solution, and suggest prevention — then assembles a single searchable solution file in docs/solutions/.

0
installs
▌
Constatar Plan · jairotorregrosa
Author plan.toml files for the constatar engine — units, dependencies, adapters, and verification plans per the 6-rung ladder. Use when the user wants to plan or orchestrate multi-agent work with constatar.

0
installs
▌
Constatar Verify · jairotorregrosa
Run and interpret verification through the constatar engine — the 6-rung ladder, grounded evidence, and conformance auditing. Use when the user asks to verify work with constatar, run a constatar plan, or audit a constatar run.

0
installs

12 published skills