Scenario 01
Technology decision research
Agent task requires synthesizing information from multiple sources into a decision framework.
Research the top 5 vector database options and give me a decision matrix for our use case.
Run Deep, Multi-Turn Agent Tasks Without Losing the Thread
Antigravity's Agent Manager is built for autonomous execution — but complex tasks like vendor evaluations, architecture reviews, and policy drafts need multiple rounds of reasoning, feedback, and revision. Felo SuperAgent adds a persistent multi-turn layer: the agent remembers what it decided, why it decided it, and where the task is heading — across as many turns as the work requires.
How it works
Agent tasks that trigger felo-superagent automatically
These are the prompts where Antigravity's Agent Manager routes to felo-superagent — when the task requires deep, multi-source research.
01 · User
Prompt sent
You type a task in Antigravity. No special syntax needed.
02 · Agent Manager
Task analysis
Gemini 3 plans the task and detects knowledge cutoff risk.
03 · SKILL.md router
Semantic match
SKILL.md description matches the task. The skill is loaded.
04 · Skill
Skill execution
The matched skill runs and fetches or processes the data.
05 · Output
Cited answer inline
Structured answer with source links returned to the agent.
Scenario 02
Competitive landscape analysis
Agent needs to build a comprehensive view of a competitive space from multiple sources.
Build a competitive landscape for AI coding assistants — features, pricing, and market position.
Scenario 03
Due diligence research
Agent task involves deep research on a vendor, technology, or market before a decision.
Do a technical due diligence on Temporal.io — architecture, limitations, and production case studies.
How it works
What felo-superagent adds to Antigravity
Six capabilities that extend Antigravity's Agent Manager for deep, multi-turn research tasks.
Multi-turn task execution
Runs deep research tasks across multiple steps without losing context or thread.
Cross-source synthesis
Aggregates findings from multiple sources into a single structured output.
Persistent task memory
Remembers intermediate findings across agent turns — no re-fetching.
Sub-task delegation
Breaks complex research into parallel sub-tasks and merges results.
Structured report output
Returns a formatted research report the agent can pass directly to you.
SKILL.md auto-routing
Activated when the agent task requires deep, multi-step research or synthesis.
Use Cases
Real tasks, real workflows
How developers use this skill inside Antigravity — from quick lookups to complex multi-step workflows.
01
Technology decision research
Agent task requires synthesizing information from multiple sources into a decision framework.
❯Research the top 5 vector database options and give me a decision matrix for our use case.
02
Competitive landscape analysis
Agent needs to build a comprehensive view of a competitive space from multiple sources.
❯Build a competitive landscape for AI coding assistants — features, pricing, and market position.
03
Due diligence research
Agent task involves deep research on a vendor, technology, or market before a decision.
❯Do a technical due diligence on Temporal.io — architecture, limitations, and production case studies.
Installation
Three steps to persistent multi-turn agent tasks in Antigravity
Once installed, felo-superagent activates for tasks that require sustained reasoning — the Agent Manager uses it when a single pass isn't enough.
Method 1 · Terminal
Install via command line
Clone the Felo skills repo and copy the skill into your global Antigravity skills directory.
# Clone the Felo skills repository
git clone https://github.com/Felo-Inc/felo-skills.git
# Copy this skill to your global Antigravity skills folder
cp -r felo-skills/felo-superAgent ~/.gemini/antigravity/skills/Installs to ~/.gemini/antigravity/skills/ — available across all your projects
Method 2 · Agent prompt
Let the agent install it for you
Paste this prompt directly into Antigravity. The agent clones the repo and copies the skill automatically.
Paste into Antigravity
Install https://github.com/Felo-Inc/felo-skills into my global Antigravity skills directory.
The agent handles cloning, copying, and directory setup — no terminal needed
1
Copy the skill into .agent/skills/
Clone the Felo skills repo and copy felo-superagent into .agent/skills/ for team-wide access. The SKILL.md description tells the Agent Manager to activate it for complex, multi-step tasks that need continuity.
2
Set your Felo API key
Add FELO_API_KEY to your environment. felo-superagent works alongside Antigravity's native Agent Manager — it adds the persistent context layer that keeps long tasks coherent.
3
Start a task, iterate without losing context
Kick off a complex task — a vendor evaluation, an architecture review, a rollout plan. The agent works through it in rounds, preserving decisions and reasoning between turns. You review, redirect, and the agent continues from exactly where it left off.
No configuration needed. The SKILL.md file inside the skill folder tells Antigravity's Agent Manager when to activate it. Once the folder is in place, the agent routes to the skill automatically based on your prompt.
FAQ
Felo SuperAgent in Google Antigravity — FAQ
Antigravity's Agent Manager is excellent at planning and executing discrete tasks. felo-superagent adds the layer for tasks that need to evolve over multiple rounds — where the agent needs to remember earlier decisions, incorporate feedback, and maintain a coherent thread across turns.
Handle complex, multi-turn tasks without losing the thread
Install felo-superagent in .agent/skills/ and let Antigravity agents work through long-horizon tasks with persistent context and accumulated reasoning.
One API key. Commit to .agent/skills/ and your agents can handle tasks that take days, not minutes — without losing context.
More Antigravity Skills
