Agent Advisor — Security skill for Claude Code
Hub for selective AI-agent routing across Codex, Claude Code & Antigravity: auditable solo/delegate/audit/full modes, canonical specs, cross-harness comparison, and fail-closed sync to three installab.
How to install Agent Advisor
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open toshon-jennings/agent-advisor and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Agent Advisor does
Hub for selective AI-agent routing across Codex, Claude Code & Antigravity: auditable solo/delegate/audit/full modes, canonical specs, cross-harness comparison, and fail-closed sync to three installable plugins.
Alternatives in Security
- Synk Skill Security Scanner — Agent Trust Hub 1.9k ★
- Engage.Cvediff — Find the canonical fix commit(s) for a CVE across sources, then diff for root cause 348 ★
- The Openclaw Optimizer — AI skill for optimizing and troubleshooting OpenClaw workspaces 67 ★
README
Agent Advisor
One routing model, three execution harnesses.
**This repo is maintainer infrastructure — there is nothing here to install.** To *use* selective routing, install the plugin for your harness:
- claude-advisor (Claude Code)
- gem-advisor (Antigravity)
- sol-advisor-portable (Codex):
- Adds 2 primary chairs (Daybreak Blue and Astra)
- Adds a one-command installer that works on any machine, stops safely instead of overwriting your setup, a check command to verify the install, short names to pick Sol / Astra / Daybreak, a /sol shortcut, and setup instructions for multiple devices.
This repo exists so those three stay honest with each other.
It is public for one reason: [`specs/`](specs/) and the [comparison matrix](#the-comparison-matrix) are worth reading whether or not you ever touch the sync tooling. They are a written-down account of what the same routing model costs to enforce on three different harnesses — including where one of them cannot enforce it at all.
Agent Advisor is the canonical hub for three standalone implementations of the same selective-routing workflow: **Sol Advisor** on Codex, **Claude Advisor** on Claude Code, and **Gem Advisor** on Google Antigravity. The routing model lives here in [`specs/`](specs/). Each harness's source lives in [`platforms/`](platforms/) and is exported outward to its own repository by [`scripts/sync-spokes.sh`](scripts/sync-spokes.sh).
The workflow originates with [Sol Advisor](https://github.com/DannyMac180/sol-advisor) by Daniel McAteer, MIT licensed. See [ORIGIN.md](ORIGIN.md) for what is carried over, what is newly built, and how close each adaptation actually is.
The routing model, in one paragraph
Before the first task tool call, the orchestrator emits a machine-auditable `SELECTIVE ROUTE` declaratio
Related Skills
Reviewgate
Fail-closed independent review loop for Claude Code and Codex. Native hooks, six reviewer paths (Codex, Gemini
Pipeline Team
Orchestrates the full implementation pipeline (impl-plan -> impl -> audit -> fix -> merge) across multiple pre
Auxi Figma Audit
3-pass Figma design audit for the Auxi RN app. Given a Figma URL + target screen, runs Pass 1 (Figma deep extr
Skill Security Web
Una skill para usar con Claude Code que cubre las tres capas de seguridad web esenciales: RLS (para que cada u
Fable Skills
Five skills Fable wrote before it left: the security sweep, the setup, the build planner, the honest advisor,
Review Test Quality
Audit test suite quality: coverage gaps, weak assertions, missing negative cases, over-mocking, and tests that
Related Agents
Copilot Agent
Runs GitHub Copilot CLI on the user's Copilot subscription. Three modes - consult (harness-enforced read-only
Scope Reviewer
Strategic scope review — challenge assumptions about WHAT to build before design work begins. Four modes: EXPA
Design Learner
Harness component 07 · learning-loop. Given one or more critic verdicts (solo design-critic or multi-critic sp