Build and execute AI agent workflows with a declarative trajectory CLI.
npx skills add https://github.com/ATaC-team/ATaC --skill atac
Please help me install this Agent Skill. Skill: atac Source: https://github.com/ATaC-team/ATaC/tree/main/skills/atac Command: npx skills add https://github.com/ATaC-team/ATaC --skill atac
ATaC provides a unified approach to building, recording, and replaying AI agent trajectories by combining declarative workflows with a CLI, reducing ad-hoc scripting overhead and improving reproducibility.
Create a new trajectory workspace with atac init demo and build a simple for-loop workflow, then run it with atac run demo.
High-intent search queries and answers about installing and using this skill.
AI agent workflows can be orchestrated using a declarative trajectory DSL that supports loops, conditionals, and tool calls across Bash and MCP environments. This enables deterministic tool orchestration and reduces ad-hoc scripting overhead.
A declarative trajectory language defines AI agent workflows structurally rather than through ad-hoc scripts. It allows you to specify loops, conditionals, and tool calls deterministically, improving the reproducibility of automated tasks.
To set up an AI workflow from the CLI, create a workspace using the init command, build a loop workflow using add-input, add-for, and add-action commands, then execute it with the run command.
Yes, you can run nested sub-workflows for scalable automation by invoking bash://run within your trajectory. This enables modular workflows and reusability while operating in a global working directory with context flags.
Yes, this workflow approach operates across both Bash commands and MCP tools. The trajectory DSL supports deterministic tool orchestration seamlessly across these environments without requiring separate scripts.
Using a declarative DSL for AI agent trajectories reduces ad-hoc scripting overhead and improves reproducibility. Unlike manual Bash scripts, it natively supports loops, conditionals, and modular sub-workflows for deterministic execution.
Bitterbot-AIList, configure, authenticate, and call MCP servers and tools via CLI.
davidahmannConvert Gait run artifacts into deterministic regression workflows with JUnit reports.
TeamSparkAIBuild no-code MCP servers using YAML graphs with JSONata and JSON Logic.
damusixWrite shell scripts and automation using Google's zx library in JavaScript/TypeScript.
