Aller au contenu
TakoAPI
Retour aux skills

lattice

par cnf6682

DevOps & Cloud
511Téléchargements rapportés par clawskills.sh
2Étoiles
0Vues
0 mention J'aimeConnecte-toi pour aimer ce skill
—Pas encore noté
Ta note

Résumé

Initialize and manage Lattice organizations — a file-based operating system for AI agent teams that enables stable.

Description

Ce que fait ce skill

Lattice is a file-based operating system for AI agent teams. It structures development work as an 8-phase pipeline (Constitute → Research → Specify → Plan → Implement → Test → Review → Gap Analysis), with each phase's state written to disk so agents can resume across sessions without context loss. When an agent gets stuck, a three-tier system handles recovery automatically: retry with a stronger model, consult multiple models in parallel, or auto-triage to relax constraints, defer, or flag for human review.

File-driven state avoids the context window overflow and drift that derail agents in chat-based workflows, letting tasks complete reliably over hours or days.

Quand l'utiliser

  • Running a multi-day refactor across agent sessions without losing progress
  • Assigning cheaper models to code-writing phases and stronger models to planning phases to cut token costs
  • Automatically unblocking a stuck agent without manual intervention
  • Running several software projects in parallel on independent cron schedules
  • Setting up a structured department and project hierarchy for a new agent team

Exemple de workflow

ENTRÉE

User asks: Set up a Lattice org for my backend API project with Engineering and QA departments

AGENT
  1. 1

    Asks for org name, target directory, department names, and optional first project details

  2. 2

    Copies templates/ORG/ to the target directory and customizes ORG_README.md with Engineering and QA departments

  3. 3

    Scaffolds department directories from the example-dept template, filling in names and resetting STATE.json

  4. 4

    Configures PIPELINE_STATE.json with per-phase model assignments and escalation chain based on user input

  5. 5

    Creates a cron job running the orchestrator prompt every 30 minutes and reports the job ID

SORTIE

ORG/ directory created at /workspace/ORG/ with Engineering and QA departments, one project configured with pipeline state, and an orchestrator cron job scheduled every 30 minutes

Prérequis

  • API keys for each AI model provider assigned to pipeline phases (e.g., Anthropic, OpenAI, Google)

Installation

clawhub install cnf6682/lattice

Exécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.

Liens