DEV Community

gracefullight
gracefullight

Posted on

oh-my-agent: cross-context reviews and trigger accuracy testing

When you ask a model to review its own design, it usually gives itself an A. We merged a fix for authorship bias this week.

What's new

  • Trigger accuracy harness: oma verify triggers now measures keyword false-fire and missed-fire rates against a 167-entry labeled corpus with zero LLM calls.
  • Standards emission: oma emit generates conformant Agent Skills spec folders, a .claude-plugin marketplace manifest, and an AGENTS.md index directly from the .agents single source of truth.
  • Skills linting: oma skills lint enforces deterministic authoring-smell checks based on Anatomy-to-Smells research.
  • Bundle sprawl warnings: oma skills audit now flags skills that exceed 20 reference files or 25,000 characters.

We also wired up SessionStart hooks for Claude Code to reload skills dynamically and handle session compaction without dropping context.

What's fixed

  • Keyword matching now ranks by specificity. Longest keyword wins, then multi-word, then declaration order. This resolves self-suppression collisions.
  • Relayed inter-agent messages no longer trigger false workflow keyword detections.
  • Windows atomicWriteJson failures (EPERM) are resolved by opening the temp file with read-write permissions.
  • The opencode bridge is completely rewritten to honor the new plugin API mutation contract, restoring true persistent mode.

What's better

  • Cross-context reviews: All 11 ultrawork review steps now dispatch to fresh reviewer subagents that receive only durable artifacts and the review guide. Same-session self-review is removed because isolated reviews catch more defects.
  • High-stakes blind review: Brainstorm workflows now escalate to independent subagents for design artifact critique, completely removing authorship bias.
  • Marketplace consolidation: The repository root .claude-plugin directory is now the single source of truth for the Claude plugin marketplace.

Installation

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/first-fluke/oh-my-agent/main/cli/install.sh | bash
Enter fullscreen mode Exit fullscreen mode
# Windows (PowerShell)
irm https://raw.githubusercontent.com/first-fluke/oh-my-agent/main/cli/install.ps1 | iex
Enter fullscreen mode Exit fullscreen mode

oh-my-agent is built for teams who orchestrate more than they prompt. Next up: native Antigravity plugin emission.

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

Top comments (0)