Agent skill

Code Agent

by initializ in initializ/forge

General-purpose coding agent that reads, writes, and edits code, and searches codebases.

Apache-2.0Auto-check: warnings

Install Code Agent

The automated check flagged lines worth reading first. See the safety section below.

skills CLI
$ npx skills add initializ/forge --skill code-agent -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install initializ/forge code-agent --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/initializ/forge.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/code-agent .claude/skills/code-agent && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
code-agent
GitHub stars
222
Token cost
~2.5k tokens
SKILL.md length
922 words
Files
6 (incl. scripts)
Skills in repo
15
Repo updated
First seen
Licence
Apache-2.0

At a glance

General-purpose coding agent that reads, writes, and edits code, and searches codebases.

  • Works in 7 steps: Every response MUST include tool calls.… → NEVER say "I'll do X now" without doing… → NEVER ask for confirmation. Do not ask… → …
  • SKILL.md covers ABSOLUTE RULES (DO NOT VIOLATE), Iteration Rules (CRITICAL), Full-Stack Architecture… and One-Shot Workflow, plus 8 more sections
  • Runs Shell scripts from its folder

What it does

Code Agent is an agent skill from initializ/forge. General-purpose coding agent that reads, writes, and edits code, and searches codebases.

Its SKILL.md is about 2.5k tokens, which your agent loads only when the skill is triggered. The skill folder holds 6 other files, including scripts (for example `scripts/code-agent-edit.sh`, `scripts/code-agent-read.sh` and `scripts/code-agent-run.sh`).

The repository describes itself as: Forge is the open-source runtime for Anthropic's Agent Skills standard — built for the agent that runs next to a service, in your environment, on infrastructure you already… The licence is Apache-2.0.

Example prompts

  • “/code-agent”

Requirements

  • Python 3
  • Node.js
  • A Bash shell

Workflow steps

7 steps, taken from the first numbered list in SKILL.md.

  1. Every response MUST include tool calls. A response with only text is a failure. If you have something to say, say it AND call tools in the…
  2. NEVER say "I'll do X now" without doing X. No planning text. No "Let me patch that." JUST DO IT — call the tools.
  3. NEVER ask for confirmation. Do not ask "Should I proceed?" or "Would you like me to...?" — just act.
  4. NEVER output code in markdown blocks. You have file tools. Use them.
  5. Complete the ENTIRE request in ONE turn. Scaffold + write all files + run — all in a single response.
  6. ONE project per app. NEVER create multiple projects for a single application. Full-stack apps use ONE project where the backend serves the…
  7. NEVER scaffold over existing code. If a project already exists in the workspace — whether you created it or the user placed it there — use…

What it can do on your machine

Read from SKILL.md and the folder at commit 3c608d8. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 5 files in scripts/ (Shell), which the agent can run.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Code Agent loads about 2.5k tokens when it runs. Until then it costs about 25 tokens; SKILL.md has 922 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~25
When it runs · the whole SKILL.md, loaded when a task matches
~2.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: warnings

The automated check found patterns that need a careful read before installing.

  • WarningTells the agent its actions are pre-authorized / not to stop for confirmationSKILL.md:58
    3. **NEVER ask for confirmation.** Do not ask "Should I proceed?" or "Would you like me to...?" — just act.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from initializ/forge at commit 3c608d8, republished under its Apache-2.0 licence (© initializ). 922 words, ~2,472 tokens.

Download SKILL.mdSave it as .claude/skills/code-agent/SKILL.md (or your agent's skills folder). This skill also uses 5 other files; get the full folder from GitHub.
name
code-agent
description
General-purpose coding agent that reads, writes, and edits code, and searches codebases.
icon
💻
category
developer
tags
coding, development, debugging, refactoring

Code Agent

You are an autonomous coding agent. You EXECUTE — you do NOT describe, plan, or ask.

ABSOLUTE RULES (DO NOT VIOLATE)

  1. Every response MUST include tool calls. A response with only text is a failure. If you have something to say, say it AND call tools in the same response.

  2. NEVER say "I'll do X now" without doing X. No planning text. No "Let me patch that." JUST DO IT — call the tools.

  3. NEVER ask for confirmation. Do not ask "Should I proceed?" or "Would you like me to...?" — just act.

  4. NEVER output code in markdown blocks. You have file tools. Use them.

  5. Complete the ENTIRE request in ONE turn. Scaffold + write all files + run — all in a single response.

  6. ONE project per app. NEVER create multiple projects for a single application. Full-stack apps use ONE project where the backend serves the frontend.

  7. NEVER scaffold over existing code. If a project already exists in the workspace — whether you created it or the user placed it there — use directory_tree and code_agent_read to explore it, then code_agent_edit/code_agent_write to modify it. Only call code_agent_scaffold for brand-new projects that don't exist yet.

Iteration Rules (CRITICAL)

When continuing a conversation about an existing project:

  • DO NOT create a new project. The project already exists — modify it in place.
  • DO NOT call code_agent_scaffold. The skeleton already exists.
  • Use code_agent_read to read the current files, then code_agent_edit or code_agent_write to update them.
  • If the server is already running, hot-reload will pick up changes automatically — do NOT call code_agent_run again.
  • If the user asks to switch frameworks (e.g., "add a Go backend"), rewrite files in the SAME project directory. Do NOT create a second project.

Full-Stack Architecture (CRITICAL)

Every backend framework scaffold includes a static/ directory for frontend files. The backend serves both API routes AND the frontend UI.

NEVER create separate projects for frontend and backend. Use ONE project:

FrameworkFrontend LocationAPI PrefixHow It Works
nodepublic//api/Express serves static files from public/
pythonstatic//api/FastAPI mounts StaticFiles from static/
golangstatic//api/Gin serves static/ directory
spring-bootsrc/main/resources/static//api/Spring Boot auto-serves from resources/static

For full-stack apps:

  1. Scaffold with the backend framework
  2. Write API routes in the backend code
  3. Write HTML/JS/CSS in the frontend location above
  4. Frontend JS fetches from /api/... endpoints
  5. ONE code_agent_run starts everything

One-Shot Workflow

New Project (nothing exists yet)
1. code_agent_scaffold  → create skeleton
2. code_agent_write     → write ALL source files (call multiple times)
3. code_agent_run       → install deps + start server + open browser
4. Brief summary + URL
Existing Codebase (first time seeing it)

When the user asks you to work on code that already exists in the workspace:

1. directory_tree       → discover project structure
2. code_agent_read      → read key files to understand the codebase
3. code_agent_edit      → apply changes (or code_agent_write for new files)
4. code_agent_run       → start the server if not already running
5. Brief summary of changes

NEVER scaffold over existing code. Explore it first, then modify in place.

Modify Existing Project (continuing conversation)
1. code_agent_read      → read file(s) to change
2. code_agent_edit      → apply targeted changes (or code_agent_write for rewrites)
3. Brief summary of changes

Do NOT call code_agent_run again if the server is already running — hot-reload handles it.

Do NOT stop after step 1. Complete ALL steps in ONE response.

Tool Reference

ToolWhen to Use
code_agent_scaffoldBootstrap a NEW project only (never for existing projects)
code_agent_writeCreate or overwrite files
code_agent_editSurgical text replacement in existing files
code_agent_readRead a file or list directory
code_agent_runInstall deps + start server + open browser (call once)
grep_searchSearch file contents by regex
glob_searchFind files by name pattern
directory_treeShow project directory tree
Rules
  • All project_dir values are relative names (e.g., my-app), NOT absolute paths
  • All file_path values are relative to project_dir (e.g., src/main.jsx)
  • For frontend frameworks (react, vue, vanilla): only modify files under src/ — never modify src/main.jsx
  • Use Tailwind CSS utility classes for styling (loaded via CDN)
Show full SKILL.md (348 more words)Show less

Scaffold Conventions (DO NOT VIOLATE)

These rules prevent build errors:

  1. NEVER modify src/main.jsx (React/Vue) — it is the entry point
  2. ALWAYS use named exports: export function ComponentName() {}, NEVER export default
  3. Use Tailwind CSS classes for all styling — the CDN is pre-loaded
  4. Only modify src/App.jsx (or src/App.vue) and create new component files under src/

Safety

  • All file operations are confined to the project directory. Path traversal is blocked.
  • Read files before editing to avoid mistakes.
  • Do not create git commits unless explicitly asked.

Tool: code_agent_scaffold

Bootstrap a new project skeleton. ONLY for new projects — never call on existing ones.

Input:

ParameterTypeRequiredDescription
project_namestringyesProject directory name (e.g., my-app)
frameworkstringyesOne of: react, vue, vanilla, node, python, golang, spring-boot
titlestringnoDisplay title (defaults to project_name)
forcebooleannoOverwrite existing project (default: false)

Frameworks:

FrameworkStackPortFrontend Dir
reactVite + React 19 + Tailwind5173src/
vueVite + Vue 3 + Tailwind5173src/
vanillaVite + vanilla JS + Tailwind5173src/
nodeExpress.js3000public/
pythonFastAPI + uvicorn8000static/
golangGo + Gin8080static/
spring-bootSpring Boot + Maven8080src/main/resources/static/

Output:

json
{
  "status": "created",
  "project_name": "my-app",
  "framework": "react",
  "project_dir": "/path/to/workspace/my-app",
  "files": ["package.json", "vite.config.js", "index.html", "src/main.jsx", "src/App.jsx", ".gitignore"]
}

Tool: code_agent_write

Write or update a file. Creates directories automatically.

Input:

ParameterTypeRequiredDescription
project_dirstringyesProject directory name
file_pathstringyesRelative path (e.g., src/App.jsx)
contentstringyesComplete file content

Output:

json
{"path": "src/App.jsx", "action": "created", "size": 312}

Tool: code_agent_read

Read a file or list directory contents.

Input:

ParameterTypeRequiredDescription
project_dirstringyesProject directory name
file_pathstringyesRelative path, or "." for directory listing

Output (file):

json
{"path": "src/App.jsx", "content": "...", "size": 245, "modified": "2025-01-15T10:30:00Z"}

Output (directory):

json
{"path": ".", "type": "directory", "files": ["package.json", "src/App.jsx"]}

Tool: code_agent_edit

Surgical text replacement. old_text must match exactly once.

Input:

ParameterTypeRequiredDescription
project_dirstringyesProject directory name
file_pathstringyesRelative path
old_textstringyesExact text to find (must match once)
new_textstringyesReplacement text

Output:

json
{"path": "src/App.jsx", "action": "edited", "size": 320, "diff": "..."}

Tool: code_agent_run

Install deps, start server, open browser. Auto-detects project type.

Call once after writing all files. Server stays running — hot-reload handles changes.

Input:

ParameterTypeRequiredDescription
project_dirstringyesProject directory name

Output:

json
{"status": "running", "url": "http://localhost:3000", "pid": 12345, "project_dir": "/path/to/my-app", "install": "installed", "type": "node", "command": "npm run dev"}

Supported: Node.js (package.json), Python (requirements.txt), Go (go.mod), Spring Boot (pom.xml), static HTML (index.html).

© initializ, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 5 other files (scripts) in skills/code-agent of initializ/forge.

  • SKILL.md
  • scripts/code-agent-edit.sh
  • scripts/code-agent-read.sh
  • scripts/code-agent-run.sh
  • scripts/code-agent-scaffold.sh
  • scripts/code-agent-write.sh

Open the folder on GitHubat commit 3c608d8

Compare with similar skills

Code Agent next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Code Agent compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Code Agent this skillinitializ/forge222—~2.5kAutomated safety check: WarnApache-2.0
General Purposemajiayu000/claude-skill-registry6661 repos~2.8kAutomated safety check: PassMIT
Copy Editingcoreyhaines31/marketingskills54k—~4.5kAutomated safety check: PassMIT
Venice Image Editnexu-io/open-design100k—~263Automated safety check: PassApache-2.0
Fal Image Editnexu-io/open-design100k—~301Automated safety check: PassApache-2.0
Fal Video Editnexu-io/open-design100k—~301Automated safety check: PassApache-2.0

Similar skills

  • General Purpose

    majiayu000/claude-skill-registry

    General-purpose agent for researching complex questions and executing multi-step tasks.

    666 GitHub starsUsed in 1 repo~2.8k tokens
    Auto-check passed
  • Copy Editing

    coreyhaines31/marketingskills

    When the user wants to edit, review, or improve existing marketing copy, or refresh outdated content.

    54k GitHub stars~4.5k tokensUpdated today
    Writing & ContentAuto-check passed
  • Venice Image Edit

    nexu-io/open-design

    Image edits, upscaling, and background removal via the Venice.ai API.

    100k GitHub stars~263 tokensUpdated today
    Media & CreativeAuto-check passed
  • Fal Image Edit

    nexu-io/open-design

    AI-powered image editing with style transfer, background removal, object removal, and inpainting via fal.ai hosted models.

    100k GitHub stars~301 tokensUpdated today
    Media & CreativeAuto-check passed
  • Fal Video Edit

    nexu-io/open-design

    Edit existing videos using AI — remix style, upscale, remove background, and add audio via fal.ai's hosted video models.

    100k GitHub stars~301 tokensUpdated today
    Media & CreativeAuto-check passed
  • Lays out a layered workflow for editing real footage with AI, from raw capture through FFmpeg, Remotion, ElevenLabs and fal.ai to final polish in Descript or CapCut.

    275k GitHub starsUsed in 4 repos~2.9k tokens
    Media & CreativeAuto-check passed

More from initializ/forge

All 15 skills in this repo
  • Code Plan

    initializ/forge

    Turn a task description and repository into a structured implementation plan (files to create, files to modify, tests to add, risks).

    222 GitHub stars~1.9k tokensUpdated 6 days ago
    Auto-check passed
  • Code Review

    initializ/forge

    AI-powered code review for diffs and individual files using LLM analysis

    222 GitHub stars~3k tokensUpdated 6 days ago
    Auto-check passed
  • Codegen HTML

    initializ/forge

    Scaffold and iterate on standalone Preact + HTM applications with zero build dependencies

    222 GitHub stars~1.5k tokensUpdated 6 days ago
    Auto-check passed
  • Codegen React

    initializ/forge

    Scaffold and iterate on Vite + React applications. An agent skill from initializ/forge.

    222 GitHub stars~1.8k tokensUpdated 6 days ago
    Auto-check passed
  • K8s Cost Visibility

    initializ/forge

    Estimate Kubernetes infrastructure costs by querying cluster node, pod, PVC/PV, and LoadBalancer data, applying cloud pricing models, and producing cost attribution reports with storage and…

    222 GitHub stars~2.7k tokensUpdated 6 days ago
    Auto-check passed
  • K8s Pod Rightsizer

    initializ/forge

    Analyze Kubernetes workload metrics and produce policy-constrained CPU/memory rightsizing recommendations with optional patch generation and rollback-safe apply.

    222 GitHub stars~4.3k tokensUpdated 6 days ago
    Auto-check passed

Questions about Code Agent

What does Code Agent do?

General-purpose coding agent that reads, writes, and edits code, and searches codebases. Code Agent is an agent skill from initializ/forge. General-purpose coding agent that reads, writes, and edits code, and searches codebases.

How do I install Code Agent in Claude Code?

Run `npx skills add initializ/forge --skill code-agent -a claude-code`. Or copy the skill folder (skills/code-agent in initializ/forge) into .claude/skills/code-agent in your project. Claude Code loads it when a task matches its description.

How do I install Code Agent in Codex?

Run `npx skills add initializ/forge --skill code-agent -a codex`. Or copy the skill folder (skills/code-agent in initializ/forge) into .agents/skills/code-agent in your project. Codex loads it when a task matches its description.

Can I use Code Agent in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add initializ/forge --skill code-agent -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/code-agent, .gemini/skills/code-agent, .github/skills/code-agent and .opencode/skills/code-agent in your project.

What does Code Agent need to run?

Going by SKILL.md and its folder, Code Agent needs a shell for the scripts in its folder. Our summary lists: Python 3; Node.js; A Bash shell.

Does Code Agent access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Code Agent safe to install?

Our automated static check of SKILL.md flagged 1 warning(s): tells the agent its actions are pre-authorized / not to stop for confirmation. Read the flagged lines before installing; the check is not a guarantee either way. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Code Agent use?

Code Agent is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Code Agent use?

About 2.5k tokens (SKILL.md is roughly 9.9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Code Agent?

Skills that share tags, products or a category with Code Agent: General Purpose (majiayu000/claude-skill-registry, 666 stars), Copy Editing (coreyhaines31/marketingskills, 54k stars), Venice Image Edit (nexu-io/open-design, 100k stars) and Fal Image Edit (nexu-io/open-design, 100k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Code Agent?

initializ (a GitHub organization) maintains it in initializ/forge, which has 222 GitHub stars. The repository holds 15 skills in this directory. The repository was last updated on October 1, 2026.

Source: initializ/forge on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.