oh-my-agent: strict commit attribution and capability routing

When you tell an agent to write code, getting it to commit with the right attribution is surprisingly brittle. A typo in a co-author address survives a history rewrite and ruins contributor graphs forever. This week, we fixed that and tightened up how capabilities are routed across the system.

What's new

• Added a native commit-msg hook to enforce exact Co-authored-by addresses, guaranteeing correct GitHub attribution for agent commits.
• Granted oma and ls blanket permissions for Claude and Qwen, cutting down on repetitive confirmation prompts during long runs.
• Updated Ralph to lock criteria autonomously without stopping for user confirmation in persistent mode.

The git hook checks against your allowed addresses before the commit writes to disk. If you already use husky or lefthook, it politely provides the one line you need to add instead of taking over your configuration.

What's fixed

• Pointed the default serena configuration at the shared bridge daemon, saving memory by preventing duplicate server processes per session.
• Included pi and opencode extension vendors in the default linking path so they generate and document correctly out of the box.
• Replaced bare trigger words with multi-word phrases for French and Portuguese queries to eliminate false-positive workflow activations.

What's better

• Renamed 5 actor-style skill identifiers to capability nouns for clearer routing. Migration 023 handles the transition automatically.
• Upgraded the keyword detector to safely ignore multiline bug and error inquiries, keeping memory coordination focused on actual state changes.

Installation

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/first-fluke/oh-my-agent/main/cli/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/first-fluke/oh-my-agent/main/cli/install.ps1 | iex

oh-my-agent is built for teams who orchestrate more than they prompt. Next up: expanding capability discovery across distributed teams.

https://github.com/first-fluke/oh-my-agent

Comments

[Reid Marlow]

The commit-msg hook is the part I would steal first. Agent commits are easy to clean up until one bad co-author line lands in a branch people have already pulled. I like that the hook refuses the commit before Git writes the mistake into history.