on-device-delegate — GitHub Copilot skill: what it does
on-device-delegate — a GitHub Copilot skill that delegates heavy file reading to an on-device worker. Keeps the main cloud context lean during log analysis and test triage.
MIT ★ 41,430 updated 2026-10-23
gh skill install github/awesome-copilot skills/on-device-delegateWhat on-device-delegate does
The skill offloads reading tasks to the local-coding:on-device-worker agent running the MAI Code 1.1 Flash model. By processing raw text on the local machine, large logs and file trees never enter the main cloud context window. The primary session sends a structured brief, capping tool calls at eight, and receives a single structured response.
The cloud agent performs at most one quick verification on the return value, checking output format or spot-checking a count. It does not re-read the input or recalculate the entire answer. Security reviews and architecture choices must stay with the primary model. If the worker fails or reports UNABLE, the main session completes the task directly.
Operation requires Windows, GitHub Copilot CLI with experimental features enabled, the local-coding plugin, and the MAI Code 1.1 Flash model. Local execution takes minutes rather than seconds while the model spins up.
What it helps with
- keep cloud context free when reading over 25 KB of prose
- aggregate error counts across log directories locally
- summarize multiple files without loading them into main chat
- extract facts from large test reports on-device
Example prompts for the agent
- Use on-device-delegate to list every distinct error with its count from logs/
- Run on-device-delegate to extract failed assertions from test-output/
- Delegate reading of docs/specs/ to summarize module configuration values
How to install on-device-delegate
1GitHub CLI
gh skill install github/awesome-copilot skills/on-device-delegate2Manually: put the files in place
.github/skills/on-device-delegate/SKILL.md1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill on-device-delegate -a claude-code2Manually: put the files in place
~/.claude/skills/on-device-delegate/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/on-device-delegate/
1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill on-device-delegate -a codex2Manually: put the files in place
.agents/skills/on-device-delegate/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill on-device-delegate -a cursor2Manually: put the files in place
.cursor/skills/on-device-delegate/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill on-device-delegate -a gemini-cli2Platform CLI
gemini skills install https://github.com/github/awesome-copilot --path skills/on-device-delegateManually: put the files in place
.devin/skills/on-device-delegate/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about on-device-delegate
How to install on-device-delegate in GitHub Copilot?
Run gh skill install github/awesome-copilot skills/on-device-delegate. It installs the skill file to .github/skills/on-device-delegate/SKILL.md.
What does on-device-delegate do?
It briefs bounded, read-intensive sub-tasks to an on-device worker running MAI Code 1.1 Flash, preventing large text from bloating cloud context.
What is required to run on-device-delegate?
It requires Windows, GitHub Copilot CLI with experimental mode enabled, the local-coding plugin, and MAI Code 1.1 Flash installed via /model.
Source text
SKILL.md · MIT
---
name: on-device-delegate
description: 'Saves cloud-model context by briefing one bounded, read-heavy sub-task (more than about 25 KB of prose, or many files that no single command can answer) to the local-coding plugin''s on-device worker, which runs on the local MAI Code 1.1 Flash model. Use for log and test triage, summarizing many files, or extracting facts from large inputs. When the user names this skill, call the skill tool in the same response as your first other tool call, never as a separate step.'
compatibility: 'GitHub Copilot CLI on Windows with experimental features on, MAI Code 1.1 Flash installed from /model, and the local-coding plugin installed (it provides the on-device worker agent).'
---
# On-Device Delegate
Send the bulky reading in a task to the on-device worker, then check its answer
once. The main session keeps planning and answering. The worker reads the big
inputs on the local model, so they never enter the main context.
## Steps
1. **Gate on what you would read.** If one command or a short script can
compute the answer, do it yourself now: no plan, no notes about this
skill. Delegate only when you would otherwise READ more than about 25 KB
of text into your context (prose, many files). Never delegate security
review or design decisions.
2. **Check the worker.** The worker is the custom agent
`local-coding:on-device-worker` (the plugin id). If the task tool lists it
as plain `on-device-worker` instead (copied into an agents folder), use
that. If neither is available, work normally.
3. **Brief.** Do not read the big input first. Make ONE task call:
- `name`: "on-device-<short-kebab>"
- `description`: 3-5 words
- `agent_type`: the worker id from step 2
- `mode`: "sync"
- `model`: the on-device model id named in the session's routing line,
exactly (if there is none:
"winml-azureml-mai-code-1.1-flash-on-device/MAI-Code-1.1-Flash-On-Device").
The worker does not pin a model, so without this it runs on your own
model.
- `prompt`, exactly:
```text
GOAL: <one sentence>
INPUTS: <exact paths>
RULES: Read only INPUTS. At most 8 tool calls. Edit nothing but OUTPUT. No questions.
OUTPUT: <file path and exact format, or: reply only>
ANSWER: Reply ONCE with only <exact format>, then stop.
If you cannot, reply only: UNABLE: <reason>
```
4. **Check once.** At most ONE cheap check (format, a count, or one spot
check). Never recompute the answer or re-read the input. If the answer is
wrong, UNABLE, or the task call fails: do it yourself. Then finish.
## Example
User: "Use the on-device-delegate skill. From the logs under `logs\`, list
every distinct error message with its count, most frequent first."
Task prompt:
```text
GOAL: Count each distinct error message in the logs.
INPUTS: logs\
RULES: Read only INPUTS. At most 8 tool calls. Edit nothing but OUTPUT. No questions.
OUTPUT: reply only
ANSWER: Reply ONCE with only lines of the form "<count> <message>", most frequent first, then stop.
If you cannot, reply only: UNABLE: <reason>
```
Check: confirm the lines match the format and spot-check one count with a
single search. Then answer the user.
## Limits
- The worker can return a confident but wrong answer. The single check
catches format errors, not every wrong fact. Tell the user what you checked.
- Expect minutes, not seconds, per delegated task, especially on the first
call while the local model loads.
## Setup scripts
[`scripts/`](scripts/) holds two setup scripts for the user, described in the
local-coding plugin README. They are not part of a task: never run them while
you work on the user's request.
- [`scripts/Set-LocalOutputCap.ps1`](scripts/Set-LocalOutputCap.ps1) caps the
local model's output tokens in the user's `providers.json`.
- [`scripts/Update-LocalCodingPlugin.ps1`](scripts/Update-LocalCodingPlugin.ps1)
installs or updates the plugin, then runs the cap script.
© github · MIT · github/awesome-copilot