Jules Agent SDK

VérifiéSûr

Interagir avec l'agent IA Jules via son SDK Python officiel. Gérer les sessions, surveiller les activités, approuver les plans et intégrer les sources GitHub. Gère l'authentification, le polling et la reprise de session automatiquement.

Spar Skills Guide Bot
DeveloppementIntermédiaire
4002/06/2026
Claude Code
#jules#sdk#api#automation#agent

Recommandé pour

Notre avis

Ce skill permet d'interagir avec l'API Jules en utilisant le SDK Python officiel, gérant sessions, activités et approbations de plans.

Points forts

  • Gestion complète des sessions Jules
  • Intégration avec des sources GitHub
  • Support du mode automatisation et de l'approbation de plans

Limites

  • Nécessite une installation préalable du SDK
  • Workaround nécessaire pour automationMode
  • Gestion des 404 transitoires lors du polling
Quand l'utiliser

Utilisez ce skill pour automatiser des tâches via l'agent Jules, comme la création de sessions avec validation de code.

Quand l'éviter

Évitez de l'utiliser si vous n'avez pas accès à l'API Jules ou si vous préférez une interface CLI directe.

Analyse de sécurité

Sûr
Score qualité88/100

The skill installs an official Python SDK and uses it to interact with a legitimate API. It includes no destructive commands, exfiltration, or obfuscation. The workaround to use an internal client is a standard JSON POST, not a security risk.

Aucun point d'attention détecté

Exemples

Create a new Jules session
Use the Jules SDK to create a new session with automation mode AUTO_CREATE_PR, set source to my-repo on main branch, and require plan approval. Wait for completion.
List active sessions
List all active Jules sessions for the repository 'my-repo' using the SDK, then resume monitoring the first one found.
Handle session activities with polling
Use the Jules SDK to poll for activities of a session, handling transient 404 errors by retrying.

name: jules-sdk description: A skill for interacting with the Jules API using the jules-agent-sdk Python library. allowed-tools: [run_command, write_to_file, read_url_content]

Jules Agent SDK Skill

Use this skill to automate tasks using the Jules AI agent through its official Python SDK.

Core Capabilities

  • Session Management: Create, list, resumed, and wait for Jules sessions.
  • Activity Monitoring: Track session progress and retrieve agent/user messages.
  • Plan Approval: Automatically or manually approve generated plans.
  • Source Integration: List and filter available code sources (GitHub repositories).

SDK Installation

Ensure dependencies are installed:

pip install jules-agent-sdk python-dotenv

Attribute Handling (CRITICAL)

The SDK returns Pydantic-like models. Use snake_case attributes, NOT camelCase or dictionary keys.

  • session.source_context
  • session.update_time
  • session.state
  • session.sourceContext
  • session.get("state")

automationMode Workaround

The high-level client.sessions.create method may miss the automationMode field. Use the internal client to send a raw request:

from jules_agent_sdk.models import Session

data = {
    "prompt": "Your prompt",
    "sourceContext": {
        "source": "sources/github/owner/repo",
        "githubRepoContext": {"startingBranch": "main"}
    },
    "automationMode": "AUTO_CREATE_PR",
    "title": "Session Title",
    "requirePlanApproval": True
}

# Access the internal client to POST raw JSON
response = client.sessions.client.post("sessions", json=data)
session = Session.from_dict(response)

Robust Polling (404 Handling)

Backend resources like activities may return transient 404s immediately after session creation. Always wrap polling in a retry loop:

import time
from jules_agent_sdk.exceptions import JulesAPIError

try:
    activities = client.activities.list_all(session_id)
except JulesAPIError as e:
    if "404" in str(e):
        # Ignore transient 404 and continue polling
        pass
    else:
        raise e

Resume Logic

Before creating a new session, check for active ones to avoid duplicates.

# List sessions for a specific repo
resp = client.sessions.list(page_size=100)
sessions = resp.get("sessions", [])

# Filter by source and status
active_sessions = [
    s for s in sessions 
    if s.source_context and "my-repo" in s.source_context.source
    and s.state not in ["COMPLETED", "FAILED"]
]

if active_sessions:
    session_id = active_sessions[0].id
    # Resume monitoring...
Skills similaires