Deep Code Reader skill
Reads a repository module by module and builds knowledge skills, verified by a closed-book exam run by independent agents.
- 123 ★ on the repository
- Last commit
Extended access
Why this level
- Runs several subagents and spends many subscription tokens
- Clones a repository and creates files in the skills directory
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/deep-code-readerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs several subagents and spends many subscription tokens
- Clones a repository and creates files in the skills directory
This is third-party code. Review the repository files before installing.
What it does
Deep Code Reader scans a repository, finds modules and dependencies, then asks which parts to read deeply. Each module becomes a skill covering purpose, design logic, data structures, state flow and a guide to what to change when modifying it. Verification follows an ABC scheme. Agent A writes the skill, agent B reads the code without the skill and drafts questions with answer keys and source references, agent C answers while seeing only the skill. A module that fails the exam is repaired, with at most three rounds. A global index skill is built at the end, and you can ask questions about the code through it.
Who it is for. Developers entering an unfamiliar codebase who want the agent to answer precisely instead of retelling a summary.
Good fit when
- The agent answers about a repository in generalities, but you need signatures and how specific modules work.
- A team adopts a large library and wants its knowledge kept as skills for every session.
- The subscription sits mostly unused overnight, so a long repository analysis can run then.
- Before changing someone else's module you want a map of what to edit to get a result.
Not a fit when
- The repository is small and the agent copes by reading files directly.
- The agent cannot dispatch subagents or read the file system.
- The token limit is tight, since a three-agent scheme with repeated exams spends a lot.
Example request
Run `/deep-code-read ./path/to/project ~/.claude/skills/`, confirm the version and pick the routing and auth modules.Limitations
Analysis spends many tokens: each module goes through reading, an exam, a second exam and possible repair. Passing the exam confirms coverage of the tested questions on a specific source version, not complete understanding of the code. Runs isolated only at the prompt level are marked `partial_validated`, while `verified` requires scoped access or checked access logs. For a URL the repository is cloned into the skills folder. The author tested the setup with Claude Code, and other agents are described as likely compatible.
How to disable. Delete the `deep-code-read` folder and the generated `<project>-dr*` skills from the agent's skills directory, along with the cloned source repository.
FAQ
How many times does the skill ask the user?
Twice: which version (tag or branch) to take and which modules to read deeply. The rest runs automatically.
What comes out?
An index skill `<project>-dr` and module skills `<project>-dr-<module>` with a `SKILL.md` file and optionally `reference.md`.
How is skill quality checked?
By a closed-book exam: agent B writes questions about the code and agent C answers from the skill alone. A module passes a regression and a final exam.
What if the exam fails?
The skill is repaired and retested against the accumulated question bank, in at most three rounds. Unresolved gaps stay visible in the report.
Does it work with a local repository?
Yes, pass a path instead of a URL, for example `./path/to/project`.
Discussion
No questions or reviews yet.
Related
SkillsA terminal coding agent tuned for the deepseek-v4 model: deep thinking, reasoning effort control, Agent Skills and MCP
A Codex skill for reviewing AI-written code: explains blocks, names files and lines, checks goals against evidence.
Two skills for coding agents: wtf audits a project before shipping, better-codex keeps the agent disciplined while working on code
A skill that explains and audits code, telling you what it does and why instead of just accepting it
A skill that runs a full refactor so agents read code more easily: it finds hidden links and needless dependencies.
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Deep Code Reader (this card) | no | 123 | ||
| Codebase to Course | no | 5,799 | codebase-to-course turns code into a teaching course; Deep Code Reader saves knowledge as skills and tests it with an exam | |
| Easy Code Reader | no | not required | 154 | easy-code-reader offers MCP access to reading sources; Deep Code Reader produces ready per-module skills |
README in short
Skills face a closed-book exam: an agent without the skill writes questions about the code and an agent without the code answers them. The result records the source version.