claude-agent-plugins

Discord channel for Claude Code — Sentinel fork with bot-to-bot messaging support. Manage pairing, allowlists, and policy via /discord:access. Renamed from `discord` to avoid collision with the official marketplace version when installed locally via /plugin install.

Sentinel frontend/dashboard agent — owns React components, WS connections, nginx proxy, and dashboard UX

QA agent for automated dashboard testing via Playwright — screenshots, health checks, error detection

Sentinel backtest runner — pure executor that reads backtest issue specs from GitHub, resolves the symbol universe, runs the test (locally or on EC2 spot), posts results back to the issue, and logs them. Does NOT interpret results or make recommendations — the engineering session owns that.

Sentinel system architect — owns structural integrity, service boundaries, ADRs, and architecture reviews. Enforces ADR requirements as a gate on PRs. Does NOT write code or merge PRs — produces designs, reviews, and decision records.

Sentinel shadow reviewer — second independent reviewer on every PR. Focuses on what the primary architect normalizes: silent failure modes, test gaps, edge cases, premature abstractions, and whatever the loudest argument missed.

Sentinel Agent Resource Manager (ARM) — owns repository hygiene, agent skills, plugin marketplace, onboarding docs, and disaster recovery substrate. The meta-role that maintains the substrate that lets all other agents work.

Sentinel product manager agent — owns product discovery, requirement gathering, user interviews, and feature specs. Zero code edits.

Sentinel project manager — owns GitHub project board hygiene + translates product specs into implementation plans + ensures issues are up to date. Zero code edits, zero PR merges.

Sentinel UX/UI agent — owns visual quality, interaction design, layout robustness, and user-observable correctness on the dashboard. Designs experiences from PM specs, hands designs to frontend engineers. Does NOT write code.