developer
A software developer agent and fifteen skills for any programming language, usable from Claude Code, GitHub Copilot CLI and Codex. It builds and changes code, and it also works as a consultant: a question gets analysis, and a request to build or change gets the verified change.
| Skill | Use it for |
|---|---|
framework-practices | Framework best practices for the installed version, with deep guidance for ASP.NET Core and Spring Boot, looked up through Context7 or the official docs |
implementing-designs | Turning ADRs, designs and threat models from architect into enforced, verified code |
clean-code | Names, functions, comments, classes, errors, async code, translating ideas across languages |
reviewing-code | Reviewing diffs and pull requests, and preparing them |
refactoring | Code smells, the catalog of moves and behavior-preserving changes |
testing | What to test, test levels and doubles, legacy code, flaky tests, TDD |
debugging | Root causes, intermittent and production failures, tools by symptom |
secure-coding | Injection, access control, secrets, crypto, dependencies |
api-design | Endpoints, errors, pagination, idempotency, compatibility |
production-readiness | Logs, metrics, timeouts, retries, safe releases |
user-interfaces | Screen states, forms, accessibility, front-end behavior |
design-patterns | Whether a pattern fits, GRASP, modern equivalents, anti-patterns |
data-structures-and-algorithms | Choosing structures and algorithms, complexity, profiling |
relational-databases | Schemas, SQL, indexes, transactions, migrations |
pragmatic-practices | Requirements, estimates, Git, upgrades, documentation, teamwork |
developer is both an agent and a skill, so it still works in tools without custom agents. Use the architect agent to plan, design and consult on structure, and developer to implement the plan and consult on code. It hands new structural decisions back to architect. Framework lookups work best with the Context7 MCP server connected; without it the agent uses the official docs, or says that it couldn't check.
Install
You don't need to clone the repo.
Claude Code
claude plugin marketplace add facundomartinezvidal/developer-vault
claude plugin install developer@developer-vault
GitHub Copilot CLI
copilot plugin marketplace add facundomartinezvidal/developer-vault
copilot plugin install developer@developer-vault
Codex
codex plugin marketplace add facundomartinezvidal/developer-vault
codex plugin add developer@developer-vault
In Codex, the plugin provides the skills, and developer:developer holds the agent's instructions, so mention that skill to work in this role. Codex plugins can't ship custom agents yet. To also use it as a Codex subagent, install the plugin and add the agent file:
mkdir -p ~/.codex/agents && curl -fsSL https://raw.githubusercontent.com/facundomartinezvidal/developer-vault/main/plugins/developer/com.openai.codex/agents/developer.toml -o ~/.codex/agents/developer.toml
To get updates, run codex plugin marketplace upgrade developer-vault, or the matching marketplace update command in the other tools.
For a single session from a clone: claude --plugin-dir plugins/developer --agent developer:developer, or the same flags with copilot.
Develop
- Agent: edit
src/developer.md, then runpython3 scripts/build_agents.py. - References: those listed in
scripts/vault_manifest.pycome from the vault, viapython3 scripts/sync_vault.py <path-to-developer-knowledge>, so don't edit them by hand. The rest are hand-written, and say so at the top. - Static validation: create a virtual environment, install
requirements-dev.txt, then runpython scripts/eval_suite.py validatefrom this folder. Runpython -m unittest discover -s scripts/tests -vfor evaluation-accounting checks. - Behavioral evaluation:
python scripts/eval_suite.py prepare --output /tmp/developer-evaluation, thenblindandreport. The procedure is the same as in the architect plugin, and is described inAGENTS.md. - The scripts do not invoke or bill models. Add
--checkto the build and sync scripts in CI to catch stale output.