Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Agent Collab

ASecurity

Coordinate multiple AI agents working on the same machine and codebases via the devtools-mcp local collaboration layer: file-touch reporting hooks, advisory file claims (leases), conflict checks, and the /collab dashboard page, all linked to tracker tasks. Use when running more than one agent at once (Claude Code, Cursor, Codex), when an edit warns that another agent is on the file, when asked "who is working on what", or to claim files before a big refactor. A multi-user team collab server i...

9 stars
0 votes
0 copies
0 views
Added 9/27/2026
ai-agentsshell

Works with

claude codecursorclimcp

Security Analysis

A100/100

Scanned 9/27/2026

$npx -y skills add Ugbot/ai-grind --skill agent-collab --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Agent Collab?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Agent Collab
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/ugbot-agent-collab/badge)](https://www.skillsdirectory.com/skills/ugbot-agent-collab)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: agent-collab
description: >
  Coordinate multiple AI agents working on the same machine and codebases via
  the devtools-mcp local collaboration layer: file-touch reporting hooks,
  advisory file claims (leases), conflict checks, and the /collab dashboard
  page, all linked to tracker tasks. Use when running more than one agent at
  once (Claude Code, Cursor, Codex), when an edit warns that another agent is
  on the file, when asked "who is working on what", or to claim files before a
  big refactor. A multi-user team collab server is coming soon; this is its
  single-machine precursor.
---

# Local agent collaboration

Multiple agents on one machine share one tracker DB (`~/.devtools-mcp/tracker.db`).
Every file touch and claim lands there, so each agent, and the human at the
dashboard, can see who is working where. Nothing here blocks by default:
claims are **advisory** leases, surfaced as warnings.

> Coming soon: a multi-user **team collab server** extends this across
> machines and teammates. This local layer is its precursor, with the same data shapes,
> same tracker linkage.

## Prerequisite

The shared devtools service must be running (hooks are silent no-ops otherwise):

```powershell
.\scripts\devtools-service.ps1 start   # MCP :8000, dashboard :8765
```

## Set your identity first

Set a stable, human-readable label per agent session so the dashboard doesn't
show raw session ids, and link work to the task you're on:

```powershell
$env:DEVTOOLS_MCP_AGENT = "claude-refactor"   # who you are
$env:DEVTOOLS_MCP_TASK  = "GRIND-42"          # what you're working on (optional)
```

Hooks attach both to every touch automatically.

## Automatic touch reporting (hooks)

Installed with the plugin (`hooks/hooks.json`), or manually in `settings.json`.
Two scripts, both stdlib-only and never blocking:

- `hooks/report_touch.py` (PostToolUse on Edit/Write/MultiEdit/NotebookEdit):
  reports the touched file; if someone else is on it, a warning lands in the
  agent's context.
- `hooks/check_conflict.py` (PreToolUse, optional strict layer) checks for
  claims *before* the edit.

Config via env:

| Var | Meaning |
|---|---|
| `DEVTOOLS_MCP_COLLAB=0` | kill switch; hooks exit immediately |
| `DEVTOOLS_MCP_COLLAB_MODE` | `warn` (default) / `ask` (claimed files prompt the human) / `off` |
| `DEVTOOLS_MCP_COLLAB_URL` | service base, default `http://127.0.0.1:8765` |

Minimum-overhead install: PostToolUse only, because its response already carries conflicts.

## Explicit coordination (tracker_files tool)

```
tracker_files(action="status")                      # who's active, claims, recent touches
tracker_files(action="claim", repo=".", file="src/core.py",
              task_key="GRIND-42", ttl_minutes=30)  # advisory lease before a big refactor
tracker_files(action="conflicts", repo=".", file="src/core.py")  # who else is there
tracker_files(action="release")                     # drop all your claims when done
tracker_files(action="touch", repo=".", files=[...])# manual touch (non-hook clients)
```

Claims expire on their own (default 15 min) and are heartbeated by your own
touches, so a crashed agent never wedges a file. Another session claiming a
file you hold gets a clear error naming you and your task.

## Queries and the dashboard

- `tracker_query(view="activity")` and `view="claims"` return bounded tables.
- `http://127.0.0.1:8765/collab` shows sessions, active claims, recent touches;
  contested files (two+ sessions) are highlighted.
- Task detail pages (`/tracker/task/KEY`) show the file activity linked to that
  task via `DEVTOOLS_MCP_TASK` or `task_key=`.

## Workflow for a multi-agent session

1. Start the service; give each agent a `DEVTOOLS_MCP_AGENT` label and its
   `DEVTOOLS_MCP_TASK` key.
2. Partition work by task in the tracker (see tracker-breakdown skill).
3. Before a wide-ranging change, `claim` the hot files; release when done.
4. When a warning names another agent, check `tracker_files(action="status")`
   and either pick a different file or coordinate through task comments,
   don't race the edit.

Attribution

UgbotUgbot
View sourceSee grades on GitHubMore from Ugbot →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698621 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →