Martin Cousseau
Community@martin-cousseau · Warsaw
AI Engineer working on evaluating GenAI systems.
Agent Skills by Martin Cousseau
Showing 8 vetted skills indexed across 1 GitHub repositories.
martin-cousseauextend-agent
Implements user-requested changes to AgentOS agents and verifies them against the live container.
martin-cousseaudeploy-platform
Deploys an AgentOS platform to production using provider CLI scripts and verifies the live deployment.
martin-cousseauimprove-agent
Hardens an existing agent by probing, judging, and editing it against its own instructions.
martin-cousseaueval-and-improve
Runs the Agno eval suite, diagnoses failures, and fixes agent instructions until all cases pass.
martin-cousseaureview-and-improve
Audits repository documentation against code and auto-fixes mechanical drift before release.
martin-cousseausetup-platform
Guides fresh-clone setup of an AgentOS platform through Docker, environment, boot, and first-agent creation.
martin-cousseaucreate-evals
Authors and runs eval Case entries for agents in an AgentOS platform.
martin-cousseaucreate-agent
Creates and registers new Agno agents in an AgentOS platform with smoke testing.
Frequently Asked Questions About Martin Cousseau
FAQPage SchemaWhat tasks can I accomplish with Martin Cousseau's AgentOS skills?▼
You can create new agents, extend agents with MCP servers, knowledge bases, memory, sub-agents, or scheduled tasks, author and repair eval suites, autonomously harden agents against their stated instructions, set up the platform from a fresh clone, deploy it to production, and run repo-wide drift sweeps before public release.
Who are these AgentOS skills designed for?▼
They target engineers building and operating the AgentOS platform — particularly GenAI evaluation engineers who need to pin agent behavior down as tests, diagnose failing eval runs, register new agents in app/main.py, and verify changes against live containers before shipping.
How do I set up and run the AgentOS platform locally?▼
Use the setup-platform skill from a fresh clone: confirm Docker is installed, configure the .env file, boot the containers, prove the MCP endpoint is live, connect the AgentOS UI, and build your first agent. The deploy-platform skill later handles production rollout via the up.sh script.
How does eval creation and repair work in this AgentOS?▼
The create-evals skill maps what an agent promises, mines real sessions and eval history from Postgres, and writes Case entries in evals/cases.py. The eval-and-improve skill runs the suite via python -m evals, diagnoses every failure, fixes in-scope issues, and loops until all cases pass.
What prerequisites do these AgentOS skills require?▼
A machine with Docker for running containers, a configured .env file, access to the Postgres database storing sessions and eval history, and the AgentOS repository cloned locally. Production deployment additionally requires the provider CLI and account credentials plus completion of the JWT key step.