Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/dekaic/cocos-mcp/agents-mdgit clone --depth 1 https://github.com/dekaic/cocos-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/dekaic/cocos-mcp/agents-md)<a href="https://agentmods.dev/instructions/dekaic/cocos-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/dekaic/cocos-mcp/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01867 | $0.01867 |
| Opus 5 | $0.00933 | $0.00933 |
| Sonnet 5 | $0.00373 | $0.00373 |
| Haiku 4.5 | $0.00187 | $0.00187 |
Grade A, and why
cocos-mcp AGENTS.md scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -sS -X POST http://127.0.0.1:<mcp-port>/mcp \ How it starts
The opening of the file, as written. The whole thing — 173 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
本文是 cocos-mcp 的 agent 使用规则。只写插件级规则;项目自己的预览参数、业务本地服地址、测试账号和验证步骤写在项目文档里。
允许入口
使用本插件时,只允许从下面这些入口进入:
| 入口 | 用途 |
|---|---|
| 当前项目 MCP 实例 | scene / asset-db / preview / local 域操作 |
| 当前项目 HTTP MCP endpoint | 当前 agent 没注入 MCP tool 时的等价入口 |
cocos-mcp-cli offline 命令 |
.prefab 查询和修改、.anim 查询 |
| Playwright / Chrome | 浏览器里真实游戏页面的交互验证 |
.dev/refresh 的 restart-package |
重启本扩展代码 |
其它入口不作为本插件使用路径。项目文档可以补充项目自己的只读兜底信息,但不能覆盖本文件的入口规则。
绑定当前项目 MCP
同机可能同时打开多个 Cocos 项目。agent 必须先按项目根目录绑定 MCP 实例,再执行后续操作。
实例注册文件:
~/.cocos-mcp/editors/*.json
绑定规则:
- 扫描
~/.cocos-mcp/editors/*.json。 - 只保留
projectPath与当前项目根目录一致的记录。 - 校验
pid仍存活。 - 多个匹配时取
updatedAt最新的记录。 - 后续所有 HTTP MCP 请求都使用该记录的
url。 - 验证扩展代码变化时同时核对
runtimeBuildId、toolSchemaHash和loadedAt;工作树当前 HEAD/dirty 不能证明进程已加载新代码。
快速确认脚本:
python3 - <<'PY'
import glob
import json
import os
from pathlib import Path
def find_project_root(start):
cur = Path(start).resolve()
for item in [cur, *cur.parents]:
if (item / 'assets').is_dir() and (item / 'settings').is_dir() and (item / 'extensions').is_dir():
return str(item)
return str(cur)
PROJECT = os.environ.get('PROJECT_ROOT') or find_project_root(os.getcwd())
items = []
for path in glob.glob(os.path.expanduser('~/.cocos-mcp/editors/*.json')):
try:
with open(path, 'r', encoding='utf-8') as f:
data = json.load(f)
if os.path.realpath(data.get('projectPath', '')) != os.path.realpath(PROJECT):
continue
os.kill(int(data['pid']), 0)
items.append((data.get('updatedAt', ''), path, data))
except Exception:
pass
for _, path, data in sorted(items)[-3:]:
print(path, data.get('url'), data.get('previewUrl') or '')
PY
Tool 与 HTTP MCP
如果当前 agent 已注入 router 暴露的 MCP tool,使用带项目名前缀的 tool:
forest__preview_query_url
forest__asset_reimport
forest__preview_refresh_and_reload
如果当前 agent 没注入这些 tool,按上节解析当前项目 MCP,再直接调用 HTTP endpoint。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 173 lines · 1,867 tokens per session scan A eaa175661016
cocos-mcp AGENTS.md is an instructions file published in the GitHub repository dekaic/cocos-mcp (0 stars, last pushed 8d ago), licensed Apache-2.0. It adds 1,867 tokens to every session, about $0.0093 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
claudemon CLAUDE.md
Instructions for zamarrowski/claudemon, covering ai agent guidelines, project shape, general guidelines, general rules and guards & defensive code.
unity-code-style-guide AGENTS.md
Instructions for krogh-jacobsen/unity-code-style-guide, covering agents.md — unity 6 c, project setup — edit this block, never do these — they corrupt the project, deprecated in unity 6 and if you read nothing else.
nes-php-glfw copilot-instructions.md
Instructions for oliverearl/nes-php-glfw, covering github copilot development guidelines for nes emulator, agent document source of truth, project overview, code style & standards and documentation.
game-and-watch-retro-go-sd CLAUDE.md
Claude Code instructions for sylverb/game-and-watch-retro-go-sd, covering claude.md, what this project is, build / flash workflow, architecture and three storage tiers, one elf.
base-building-trap-defense-design-agent-skill CLAUDE.md
Instructions for dungnotnull/base-building-trap-defense-design-agent-skill, covering claude.md — skill 250: base-building-trap-defense-design, skill identity, problem this skill solves, harness flow summary and sub-skills.
trident-mcp AGENTS.md
Instructions for mordor-forge/trident-mcp, covering agents.md, what this is, build and test, single-file verification and e2e tests.