Fia Harness — Testing skill for Claude Code
Local, dependency-free spec-driven development harness for AI agents: spec-first, machine-validated state and CI enforcement of golden rules.
How to install Fia Harness
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open mcpedrogm-art/fia-harness and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Fia Harness does
Local, dependency-free spec-driven development harness for AI agents: spec-first, machine-validated state and CI enforcement of golden rules. No cloud, no telemetry. Español + English.
Alternatives in Testing
- Debug — You are tasked with helping debug issues during manual testing or implementation 10k ★
- Claude Codepro — by Max Ritter - Professional development environment for Claude Code with spec-driven workflow, TDD enforcemen 1.6k ★
- Mda — MDA Open Spec — a Markdown superset for agent-facing documents 615 ★
README
🧬 FIA HARNESS COMPLETO
**El sistema operativo para construir software con agentes de IA.** Especificación antes que código · Seguridad por diseño · Contexto mínimo en cada fase · Cero asunciones en silencio
`Python 3.8+` · `Sin dependencias externas` · `Mono o multi-agente` · `2 modos de trabajo` · `Reglas verificadas en CI` · `LLM-agnóstico` *(funciona con DeepSeek, OpenCode, Claude, GPT o el agente que uses)*
🗺️ El sistema en un vistazo
graph TD
PRD["📄 PRD.md
documento de negocio"] --> BOOT["⚙️ bootstrap.py
(Fase M0)"]
BOOT --> CTX["CONTEXT.md
resumen vivo del negocio"]
BOOT --> PROG["PROGRESS.md
fases M0-M3"]
CTX --> INT["🧠 Agente de IA
Entrevista técnica (M1)"]
INT --> SEC["SECURITY.md
AEO_GEO_SEO.md
decisiones en CONTEXT.md"]
SEC --> SPEC["SPEC.md
especificación aprobada
por el humano (M2)"]
SPEC --> PLAN["Tabla F0-Fn
en PROGRESS.md (M3)"]
PLAN --> GEN["🤖 task_generator.py"]
GEN --> TASK["TASK-F1.md
checklists de seguridad,
visibilidad y UX inyectados"]
TASK --> RUN["🛠️ Agente ejecuta
auditoría → diseño → código"]
RUN --> VAL["✅ Fase K
tests · typecheck · lint · build"]
VAL -- falla --> RUN
VAL -- pasa --> NEXT["PROGRESS.md actualizado
siguiente fase pendiente"]
NEXT --> GEN
**La idea central:** el PRD se lee una vez. A partir de ahí, cada fase del agente solo carga 4 archivos comprimidos (`CONTEXT.md` + sección de `SPEC.md` + `PROGRESS.md` + `TASK-Fx.md`). Nunca se reenvía el histórico completo. Contexto pequeño = respuestas más baratas, más rápidas y con menos deriva.
⚙️ Cómo funciona: tres motores
1️⃣ Fases de proceso (M0–M3) — *pensar antes de construir*
| Fase | Qué pasa | Entregable |
|---|---|---|
| M0 | bootstrap.py lee el PRD y activa el harness |
CONTEXT.md + PROGRESS.md + plantillas |
| M1 | El agente te entrevista: stack, BBDD, seguridad, visibilidad, UI/UX, |
Related Skills
Asktube
AskTube - An AI-powered YouTube video summarizer and QA assistant powered by Retrieval Augmented Generation (R
Claude Workflow
Claude Code plugin for spec-driven development. Takes features from idea to validated implementation using str
Ast Grep
This command helps to create and test ast-grep rules for code quality enforcement.
Claude Code Hooks Cookbook
Copy-paste Claude Code hooks that actually do something — auto-format, block dangerous commands, test gates, n
Agent OS Setup
Bootstrap the Agent OS data home + verify harness on this machine (product graph, registry, Playwright, auth c
Scipy Spec Rules
Standards from Scientific Python SPECs 0, 4, and 6. Apply when managing dependency versions, deprecation timel
Related Agents
Timps Backup Sentinel
Audit backup health: Time Machine last backup time, all local git repos with uncommitted/unpushed changes, lar
Kavach Billing
KAVACH billing/monetization specialist. Walks the full money kill chain - server-side price/entitlement enforc
Text Slop Cleaner
This agent should be invoked to remove machine sounding text and unnecessary comments. This includes rewriting