deep-research-gate

deep-research-gate is a skill for Claude Code from BackToCimaCoppi/Praxis. It costs 69 tokens per session (2,708 once invoked), scanned B, original, Apache-2.0.

A safety check for deciding when an AI agent may split work among several helper agents or start a deep research workflow. It distinguishes bounded tasks with a known number of steps from open-ended or recursive work that can keep expanding.

In plain words
What is it for?
Use it before launching parallel agents, deep research, or large workflows to check whether the work has a known limit, can recurse, or is an open investigation requiring user approval.
Why use it?
It prevents research or agent work from growing beyond the expected scope, time, or resource use, while allowing clearly limited parallel tasks to proceed.

Skill for Claude Code

Written for Claude Code: SessionStart hook event. Also seen: reads .claude/ paths; mentions Claude Code.

Good fit Use it before launching parallel agents, deep research, or large workflows to check whether the work has a known limit, can recurse, or is an open investigation requiring user approval.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/backtocimacoppi/praxis/deep-research-gate
Install

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.

Any agent
npx skills add BackToCimaCoppi/Praxis --skill deep-research-gate
Clone the repo
git clone --depth 1 https://github.com/BackToCimaCoppi/Praxis

Made for: Claude Code.

Wrote 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.

agentmods badge for deep-research-gate

README.md
[![agentmods](https://agentmods.dev/badge/skills/backtocimacoppi/praxis/deep-research-gate.svg)](https://agentmods.dev/skills/backtocimacoppi/praxis/deep-research-gate)
Your own site
<a href="https://agentmods.dev/skills/backtocimacoppi/praxis/deep-research-gate"><img src="https://agentmods.dev/badge/skills/backtocimacoppi/praxis/deep-research-gate.svg" alt="Measured on agentmods" height="20"></a>
Per session 69 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,708 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00069 $0.02708
Opus 5 $0.00034 $0.01354
Sonnet 5 $0.00014 $0.00542
Haiku 4.5 $0.00007 $0.00271

Measured 8d ago against content hash d145ec15cf10, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade B, and why

deep-research-gate scanned grade B 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 8d 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

在 `~/.claude/settings.json` 配置:
skills/deep-research-gate/SKILL.md · 117 lines

How it starts

The opening of the file, as written. The whole thing — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.

扇出安全门(deep-research-gate)

任何「要不要扇出一批子 agent / 要不要开深度研究·工作流」的时刻,先读这里、按这里走。

它管两件事,别搞混:① 该并行的有界任务放行别缩;② 会失控放大的深度研究/工作流焊死,要开得经用户同意


〇、先分清两类扇出(决策从这里开始)

危险的从来不是「子 agent 多」,是「扇出会不会失控放大」。先对号入座:

🟢 绿灯·有界并行 🔴 红灯·无界/递归/工作流
数量 已枚举、算得出上界(N 个明确子任务) 不可预知,子 agent 自己决定要不要再扩
成本 O(N) 封顶,事前算得出 失控放大,事前算不出上界
任务性质 执行已交代清楚的事(模拟 / 只读核验 / 已定的批处理) 开放式调查:自己去查清某课题、查多深由它定
递归 不递归,子 agent 是叶子 子 agent 内部还会再派 / 触发 deep-research / 跑 workflow
怎么办 直接并行,不必问、不必缩——串行才是错 走 §三 开闸流程 + 经用户同意

判据三问,任一为「是」→ 红灯:

  1. 数量算不出上界?(子 agent 自己决定要不要再派、要查多深)
  2. 会递归?(子 agent 内部还会再派 agent / 触发 deep-research / 跑 workflow)
  3. 是开放式调查?(任务是「自己去查清楚某课题」,而非「执行已交代的一件事」)

绿灯典型:「M 个配置 × N 个场景 = 一批模拟」这类——并行单元天然可枚举,靶子与流程全部冻结、零调查、不递归。正确做法是一次全部并行,硬拆成串行是纯浪费。 红灯典型:某个子 agent 为核实一个数据点,自行触发了深度研究,在内部再扇出一整层子 agent。


一、为什么需要这道门(失控是怎么发生的)

单个子 agent 之所以能把整个账号配额打穿,靠的是两层机制叠加,缺一不可:

  1. 子 agent 默认拿到全权限。 派研究/检索类子 agent 时若不显式指定类型,它默认是通用 agent——手里有完整工具集,能递归再派 agent、能自己触发深度研究 / 工作流。主线程以为自己只派了几个助手,实际上每个助手都能再开一整层。
  2. 工作流引擎负责放大。 深度研究与大规模递归 fan-out 都由动态工作流引擎驱动。一次触发就可能在内部展开成整层子 agent 与大量工具调用,token 消耗与主线程的预期完全脱钩。

再叠加第三个条件,就会从「贵」变成「静默地贵」:

  1. 若权限模式配置为跳过确认弹窗,上述触发全程无人拦截——没有弹窗、没有询问,等发现时配额已经见底。

后果形态:一次本该廉价的核查,可以在很短时间内把剩余配额打满并锁死账号,连带打断的是之后所有本可以正常进行的工作。无人值守时段触发尤其致命,因为没人来得及叫停。

根因落点要认准:炸配额的是子 agent 权力过大(能递归、能自开深度调查)+ 工作流引擎放大不是「主线程派了一批子 agent」这个动作本身。主线程并行派一批有界子 agent 从来不是问题——别从这类事件里学错教训、连累正常并行。


二、硬防护(只焊 workflow 引擎,不碰普通并行)

~/.claude/settings.json 配置:

"disableWorkflows": true,
"workflowKeywordTriggerEnabled": false
  • disableWorkflows: true —— 引擎级关掉「动态工作流」(深度研究和大规模递归 fan-out 全靠它)。这是开关、不是弹窗:所以「跳过确认」类权限模式吞不掉它、子 agent 也绕不过。触发时当场被拒、零 token 消耗。
  • workflowKeywordTriggerEnabled: false —— 关掉「关键词自动触发工作流」(默认是开的,prompt 里带上特定关键词就会升级成工作流,是隐患入口)。

作用域必须说清(否则会误伤并行):这两个开关焊死的是动态工作流引擎——即深度研究 / 大规模递归 fan-out 的总闸。它管不到、也不该管主线程普通的 Agent 并行调用:🟢 绿灯那种宽度并行根本不经过工作流引擎,这开关对它零影响。所以「开关开着」绝不是「不准并行派 agent」的理由

Read the full file on GitHub · 117 lines

Changes

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.

  1. 8d ago First seen · 117 lines · 69 tokens per session scan B d145ec15cf10

Subscribe to this mod's changes

deep-research-gate is a skill published in the GitHub repository BackToCimaCoppi/Praxis (6 stars, last pushed 13d ago), licensed Apache-2.0. It adds 69 tokens to every session and 2,708 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

neurolink-guide

Guide for using the NeuroLink SDK and CLI. Invoke when users ask how to use neurolink, integrate AI providers, add MCP tools, configure RAG, set up memory, deploy servers, or work with multimodal content. Covers SDK, CLI, providers, tools, and enterprise features.

juspay/neurolink · 65 tokens

github-commenting

How to post clean, rich, deduplicated GitHub PR review comments — suggestion blocks, multi-line anchors, markers, formatting rules. Load before posting or fixing any PR comment.

juspay/neurolink · 41 tokens

repo-conventions

NeuroLink's review standards — the critical rules to enforce, what NOT to comment on, the security bar, hot paths. Load before reviewing any change in this repository.

juspay/neurolink · 39 tokens

session-manager

A fallback skill for recording development sessions and restoring project context from an Obsidian vault, a folder of linked notes. It handles explicit resume, handoff, status, and end-of-session requests when automatic capture is unavailable.

Enakoneschniy/claude-dev-stack · 108 tokens

dev-research

A research workflow that uses NotebookLM, a tool for asking questions about a selected collection of documents, to answer development questions with citations.

Enakoneschniy/claude-dev-stack · 177 tokens

budget-continue

Show 4-option continuation prompt when session budget is low. Options: remind later (CronCreate), auto-continue locally (Desktop task), auto-continue in cloud (Cloud task), continue now. Trigger when: budget warning fires, session limit reached, context is full, user sees budget warning.

Enakoneschniy/claude-dev-stack · 64 tokens