delegate-task
Skill217 starsChanged 8 months agoMITView on GitHub
Delegate a task by forking a terminal session to a new terminal window. Use this when the user requests 'delegate' or 'create a new terminal' or 'new terminal: <command>' or 'fork session: <command>'.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- danielvm-git/bigpowerstheir own version, first seen Nov 6, 2026
- Samin12/skillstheir own version, first seen Aug 14, 2026
- Roentek/Claude_Code_Boilerplate_Frameworktheir own version, first seen May 5, 2026
- leavesprior/claude-skillstheir own version, first seen Feb 22, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- danielvm-git/bigpowers★ 288same name
- Samin12/skills★ 8same name
- Roentek/Claude_Code_Boilerplate_Framework★ 4same name
- leavesprior/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository8 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- create-farmerSkill
- daily-reviewSkill
- farmSkill
- historySkill
- newSkill
- start-second-brainSkill
- todaySkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 5 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 9
Follow the Instructions, execute the Workflow.
- Very specific: line 20
If user specifies "delegate to <name>", read .claude/reference/employees.json to get the repo path.
- Very specific: line 24
Before executing the command, run claude --help to understand the command and its options.
- Distinctive: line 25
Always use interactive mode (leave off -p flag)
- Very specific: line 27
Always run with --dangerously-skip-permissions
- Distinctive: line 31
When delegating to a different repo (employee), the agent operates in THEIR directory, not yours.
- Very specific: line 33
Capture source repo path: Run pwd to get current working directory (SOURCE_REPO)
- Distinctive: line 51
When delegating work related to an existing task or project:
- Distinctive: line 57
IF: The user requests a fork terminal with a summary.
- Very specific: line 59
Read, and REPLACE the .claude/skills/delegate/references/fork_summary_user_prompt.md with the history of the conversation between you and …
- Distinctive: line 79
Use single quotes around the full command, double quotes around the prompt.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Purpose
Fork a new terminal session to a new terminal window using one Claude Code.
Follow the Instructions, execute the Workflow.
Variables
DEFAULT_MODEL: opus HEAVY_MODEL: opus BASE_MODEL: sonnet FAST_MODEL: haiku
Employees
If user specifies "delegate to <name>", read .claude/reference/employees.json to get the repo path.
Instructions
- Before executing the command, run
claude --helpto understand the command and its options. - Always use interactive mode (leave off -p flag)
- For the --model argument, use DEFAULT_MODEL if not specified. If 'fast' is requested, use FAST_MODEL. If 'heavy' is requested, use HEAVY_MODEL.
- Always run with
--dangerously-skip-permissions
Cross-Repo Context (CRITICAL)
When delegating to a different repo (employee), the agent operates in THEIR directory, not yours. Before crafting the prompt:
- Capture source repo path: Run
pwdto get current working directory (SOURCE_REPO) - Use absolute paths in prompt for all source repo references:
- Source tasks:
{SOURCE_REPO}/tasks/... - Context files:
{SOURCE_REPO}/context/... - Output directory:
{SOURCE_REPO}/outputs/
- Source tasks:
- Instruct agent to use its skills: Include in prompt: "First, check your available skills at .claude/skills/ and use any relevant ones for this task."
- Specify output location explicitly: "Save deliverables to {SOURCE_REPO}/outputs/YYYY-MM-DD-
<slug>.md"
File Handling
- Instruct delegated agents to update existing vault files (tasks/, projects/, people/) directly—never create duplicates
- For deliverables, save to source repo's
outputs/YYYY-MM-DD-<slug>.<ext>using absolute path - Link from the reference file under
## Outputsusing the obsidian [[slug]] schema
Source Task Callback
When delegating work related to an existing task or project:
- Identify the source file (e.g.,
tasks/next-weeks-video-idea.md) - Include in the prompt: "When complete, update {absolute_path_to_source_file} with status and link to your output."
Fork Summary User Prompts
- IF: The user requests a fork terminal with a summary.
- THEN:
- Read, and REPLACE the
.claude/skills/delegate/references/fork_summary_user_prompt.mdwith the history of the conversation between you and the user so far. - Include the next users request in the
Next User Requestsection. - This will be what you pass into the PROMPT parameter of the agentic coding tool.
- IMPORTANT: To be clear, don't update the file directly, just read it, fill it out IN YOUR MEMORY and use it to craft a new prompt in the structure provided for the new fork agent.
- Let's be super clear here, the fork_summary_user_prompt.md is a template for you to fill out IN YOUR MEMORY. Once you've filled it out, pass that prompt to the agentic coding tool.
- XML Tags have been added to let you know exactly what you need to replace. You'll be replacing the
<fill in the history here>and<fill in the next user request here>sections.
- Read, and REPLACE the
- EXAMPLES:
- "fork terminal use claude code to
<xyz>summarize work so far" - "spin up a new terminal request
<xyz>using claude code include summary" - "create a new terminal to
<xyz>with claude code with summary"
- "fork terminal use claude code to
Workflow
- Understand the user's request.
- Craft a clear, complete prompt describing the task for the delegated agent.
- Build the claude command with prompt:
claude --model <MODEL> --dangerously-skip-permissions "<PROMPT>"- CRITICAL: The prompt MUST be included as a positional argument. Without it, Claude opens with no task.
- Execute fork:
- Default (same repo):
.claude/skills/delegate/scripts/fork_terminal.py '<full command>' - Different repo:
.claude/skills/delegate/scripts/fork_terminal.py '<full command>' --repo '<employee_path>' - Use single quotes around the full command, double quotes around the prompt.
- Default (same repo):
Examples
User says: "delegate research on X"
Execute: .claude/skills/delegate/scripts/fork_terminal.py 'claude --model opus --dangerously-skip-permissions "Research X and summarize findings"'
User says: "fork terminal to build feature Y"
Execute: .claude/skills/delegate/scripts/fork_terminal.py 'claude --model opus --dangerously-skip-permissions "Implement feature Y in this codebase"'
User says: "delegate to head-of-content: research video ideas"
-> Run pwd to capture SOURCE_REPO
-> Read employees.json, get path for "head-of-content"
-> Craft prompt with: (1) instruction to use .claude/skills/, (2) absolute paths to SOURCE_REPO files
Execute: .claude/skills/delegate/scripts/fork_terminal.py 'claude --model opus --dangerously-skip-permissions "First check .claude/skills/ for relevant skills. Research video ideas. Context: {SOURCE_REPO}/context/... Output to: {SOURCE_REPO}/outputs/..."' --repo '<employee_path>'
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

