jules-remote

name	jules-remote
description	Use this skill when dispatching atomic, isolated development tasks to Jules, a remote AI coding agent.

Jules Remote Agent

Jules is for atomic, isolated, non-urgent tasks where we don't need immediate results.

When to Use Jules

✅ Good Fit	❌ Bad Fit
Fix lint errors in 1-3 files	Large refactoring
Resolve type checking issues	Multi-file architecture
Add missing docstrings	Subjective design work
Write unit tests for existing code	Needs immediate results
Update import statements	Complex debugging
Simple bug fixes with clear repro	Vague requirements

Quick Reference

# List recent sessions
jules remote list --session 2>&1 | cat | head -20

# Create task with full context
jules remote new --session "Title: Fix lint errors in src/auth/

## Context
Files: src/auth/login.py, src/auth/session.py
Related: See .agent/codestyles/python.md for conventions

## Requirements
- Fix all ruff lint errors
- Ensure mypy passes with --strict
- Don't change public API

## Acceptance Criteria
- \`ruff check src/auth/\` returns 0 errors
- \`mypy src/auth/\` returns 0 errors"

# Check status
jules remote list --session 2>&1 | cat | grep "task_id"

# Pull completed task (review only)
jules remote pull --session <task_id>

# Pull and apply patch
jules remote pull --session <task_id> --apply

Task Template (Standard)

Title: {Action} {Target}

## Context
Files: {exact file paths}
Related: {reference docs or backlog items}
Current State: {what exists now}

## Requirements
- {specific requirement 1}
- {specific requirement 2}
- Don't: {anti-requirement}

## Acceptance Criteria
- {verifiable criterion 1}
- {verifiable criterion 2}

Task Templates by Type

Lint/Type Fix

Title: Fix lint errors in {path}

## Context
Files: {exact paths}
Tool: ruff / mypy / eslint

## Requirements
- Fix all errors reported by {tool}
- Preserve existing functionality
- Follow .agent/codestyles/{lang}.md conventions

## Acceptance Criteria
- `{tool command}` returns 0 errors

Unit Test Addition

Title: Add unit tests for {module}

## Context
Files: {source file}
Create: {test file path}
Framework: pytest / vitest / jest

## Requirements
- Cover all public functions
- Include edge cases
- Mock external dependencies

## Acceptance Criteria
- All tests pass with `{test command}`
- Coverage > 80% for the module

Docstring/Documentation

Title: Add docstrings to {module}

## Context
Files: {exact paths}
Style: Google / NumPy / JSDoc

## Requirements
- Document all public functions/classes
- Include parameter types and descriptions
- Add usage examples where helpful

## Acceptance Criteria
- `pydoc {module}` shows complete documentation
- No "missing docstring" lint warnings

Best Practices

1. Include exact file paths - Never be vague about locations
2. Reference conventions - Point to .agent/codestyles/ or .agent/references/
3. Verifiable acceptance - Use commands that return pass/fail
4. Atomic scope - One logical change, 1-3 files max
5. Clear anti-requirements - What NOT to do

Gotchas

• "In Progress" may be complete: Always check with jules remote pull <id>
• Pull quickly: File modifications conflict if local dev continues
• New files apply cleanly: Modifications to existing files may conflict
• Extract intent from conflicts: Don't force-apply, manually apply logic

Integration with Dev Matrix

When Jules task is created, add to matrix:

| {id} | QUEUED | P3 | easy | jules | - | - | - | @jules | {description} | {date} | {date} |

When pulled and applied, update to DONE.

Comparison with Other Dispatch

Mechanism	Urgency	Complexity	Isolation
Gemini CLI	Immediate	Low-Med	Any
Manual Task	Later	High	Any
Jules	Hours+	Low	Must be atomic

Jules = "fire and forget" for well-defined atomic work.