English | 中文

agent-query is a repository analysis skill for agents that support local skills.

This repository contains the installable skill bundle in skills/agent-query/. The bundle includes a prebuilt agent-query binary and the skill instructions that tell an agent when to use it.

Models are usually strong once they are in the right files. They are much less reliable during the first pass over a medium or large codebase.

Without a structural tool, an agent often:

• opens files at random
• guesses the architecture from names alone
• misses entrypoints and hub modules
• edits a leaf module before understanding its dependencies
• spends too much context on repo discovery instead of the actual task

agent-query gives the agent that missing first-pass map.

• An installable skill package for agents that support local skills.
• A bundled binary for fast repository analysis.
• Better repo context before editing, reviewing, or planning work.
• Faster answers to questions about structure, dependencies, symbols, hotspots, and summaries.

Clone the repository:

git clone https://github.com/huxint/agent-query.git
cd agent-query

Copy the skill into your agent's skills directory:

mkdir -p "$AGENT_SKILLS_DIR"
cp -R skills/agent-query "$AGENT_SKILLS_DIR/agent-query"

If your environment uses a Codex-compatible path, the equivalent location is:

mkdir -p ~/.codex/skills
cp -R skills/agent-query ~/.codex/skills/agent-query

The exact syntax depends on your agent.

• In Codex-style environments, invoke it as $agent-query.
• In other agents, call the installed skill by name using that agent's skill syntax.

Typical requests:

• Use agent-query to inspect this repository before making changes.
• Use agent-query to tell me which files I should read first.
• Use agent-query to summarize the architecture of this repository.
• Use agent-query to trace the dependencies around this module.

• The repository is large enough that manual file walking is slow and noisy.
• You want the agent to choose files based on structure instead of guesswork.
• You want a reusable structural layer before code edits or reviews.

This repository is meant to provide the skill package. It is not intended to be a full CLI manual.