tekkiiiii/the-agency/delegate

Claude Code and agent skill by tekkiiiii/the-agency

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Use to hand a task off with full context to a subagent that owns it end-to-end: \"delegate this\", offloading multi-step work, or keeping the main context clean. Snapshots open files, conversation, memory and decisions into a briefing; prefers a named specialist, falls back to general-purpose.

Install

npx vskill@latest install tekkiiiii/the-agency/delegate

SKILL.md

The same SKILL.md was published earlier as tekkiiiii/the-agency/delegate.

delegate

Delegates the current task to a general-purpose subagent with enough context to complete it independently. Unlike a focused specialist, this subagent has full tool access and drives the task to completion — no further prompting needed.

When to Use

  • You want a subagent to own a task end-to-end
  • You need to offload work without losing context
  • A focused skill agent isn't the right tool shape — the task spans files, tools, or domains

Workflow

Step 1 — Detect the Task File

The user has a file open in the IDE. Read it first — it's the anchor for context.

Read the file from `ide_opened_file` (already provided in the prompt as a system-reminder). Extract:

  • The file path
  • Any relevant content (look for task descriptions, TODOs, comments about what needs doing)
Step 2 — Find Project Memory

Determine the project directory. Common patterns:

  • ~/.claude/agents/{slug}/ — Agent/PD project
  • Current working directory or parent CLAUDE.md

Read the following files if they exist (in order of priority):

memory/next-session.md ← session-start briefing, carries forward context
memory/decisions.md ← decisions made so far
memory/heartbeat.md ← recent status / what's been worked on
memory/state.md ← machine-readable state snapshot
memory/TODOS.md ← outstanding tasks
CLAUDE.md ← project-level instructions and constraints
PROJECT.md ← project overview and goals
Step 3 — Synthesize the Briefing

Construct a briefing block that answers:

## Task
[What needs to be done — one sentence, then expanded]
## What's Been Done
[Decisions made, code written, files modified — be specific with paths]
## What's Left
[Outstanding tasks, TODOs, open questions]
## Context & Constraints
- Working file: [path]
- Project: [name/slug]
- Any constraints from CLAUDE.md or decisions.md
- Any relevant file paths the subagent needs to look at
## Verify Before Claiming Done
[The standard this task must meet — e.g., tests pass, builds, user confirms]

Be specific. "The app needs auth" is vague. "Add JWT session middleware to the /api/auth route using the existing AuthService — see src/auth/middleware.ts as a reference" is actionable.

If no project memory files exist, fall back to a brief summary based on the conversation history up to this point. State what's clear and what the subagent should infer.

Step 4 — Spawn the Subagent

Use the Agent tool — check Agency catalog first. If a named specialist fits the task domain, use it. If not, use subagent_type: general-purpose (this skill's purpose is end-to-end delegation, which general-purpose is suited for).

json
{
"description": "Complete: [1-line task summary]",
"name": "[short-slug-for-task]",
"prompt": "[full briefing block from Step 3]",
"subagent_type": "[named specialist — check Agency catalog first; general-purpose if no match]"
}

Use run_in_background: true for long-running tasks, false (default) for tasks that return a result to report back.

Step 5 — Report

Report back to the user:

  • What the subagent is doing
  • Its name (for reference if the user wants to message it directly)
  • That it will drive to completion independently

Briefing Quality Checklist

  • [ ] Working file path included
  • [ ] Task goal is explicit and unambiguous
  • [ ] What's been done is specific (file paths, not just "work in progress")
  • [ ] What's left is a clear list, not vague
  • [ ] Any constraints from CLAUDE.md are called out
  • [ ] Verification criteria are stated

Edge Cases

SituationBehavior
No memory files existSynthesize from conversation history; note uncertainty
ide_opened_file is blank/unclearInfer task from conversation; note assumption
Task is trivial (1-2 steps)Still delegate — use brief briefing, keeps context clean
Subagent spawn failsReport error; suggest running task directly
User interruptsNo special handling — subagent continues in background

Notes

  • The subagent does not report back to you automatically — it completes and the user sees the result.
  • If you need to send a message to the running subagent, use SendMessage({to: "[name]"}).
  • For running background tasks, you will be notified when it completes.
  • Do NOT send a shutdown message to the subagent unless the user explicitly requests it.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
3
7-day trend score
7

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/tekkiiiii/the-agency/delegate/badge)](https://verified-skill.com/skills/tekkiiiii/the-agency/delegate)

Something wrong with this skill or its page? Tell us.