🍳

Agent Recipes

Pre-built workflows combining real capabilities for common agent tasks. Each recipe shows the capability stack, risk profile, and starter code.

📦 6 recipes·Updated as capabilities grow·API docs →

File-Watching Agent

🟡 Caution

Monitor a directory for changes and trigger downstream actions — notifications, summaries, or backups — whenever files are created or modified.

DevOps automation, document processing pipelines

Capability Stack

Read FileMCP🟡
List DirectoryMCP🟢
Shell ExecOpenClaw🔴
pythonSDK docs →
from agentsift import AgentSift

client = AgentSift()

# Find the right capabilities
caps = client.capabilities.list(
    platform="mcp",
    category="file-system",
    risk="safe",
)
print([c.name for c in caps])
#filesystem#automation#monitoring

Email + Calendar Assistant

🟡 Caution

Read incoming emails, extract action items, and create calendar events or task entries automatically. The classic "inbox zero" agent.

Personal productivity, executive assistants

Capability Stack

Gmail ReadComposio🟢
Gmail SendComposio🟡
Calendar CreateComposio🟡
typescriptSDK docs →
import { AgentSift } from 'agentsift';

const client = new AgentSift();

// Find email + calendar capabilities
const caps = await client.capabilities.list({
  platform: 'composio',
  q: 'gmail calendar',
});

// Compare read vs write risk profiles
const comparison = await client.compare(
  'composio-gmail-read',
  'composio-gmail-send',
);
console.log(comparison.permissionMatrix.highRisk);
#email#calendar#productivity

Code Review Bot

🟡 Caution

Fetch a pull request diff, run static analysis, and post a structured review comment. Pair with a code execution sandbox for test-running.

CI/CD pipelines, developer tooling

Capability Stack

GitHub PRComposio🟢
Shell ExecOpenClaw🔴
GitHub CommentComposio🟡
pythonSDK docs →
from agentsift import AgentSift

client = AgentSift()

# Semantic search for code review capabilities
results = client.search.semantic(
    "review pull request and post comment",
    limit=10,
)
for r in results:
    print(f"{r.name} ({r.platform}) — score: {r.score:.3f}")
#github#code-review#devtools

Web Research Pipeline

🟢 Safe

Search the web, scrape pages, summarize content, and store structured results. Build a research assistant that turns a topic into a briefing doc.

Market research, competitive intelligence, content creation

Capability Stack

Brave SearchMCP🟢
Web FetchOpenClaw🟢
Write FileMCP🟡
typescriptSDK docs →
import { AgentSift } from 'agentsift';

const client = new AgentSift();

// Find all safe web capabilities
const webCaps = await client.capabilities.list({
  category: 'web-browsing',
  risk: 'safe',
  limit: 20,
});

// Get the full picture on a specific capability
const search = await client.capabilities.get('mcp-brave-search');
console.log(search.permissions, search.trustScore);
#web#research#scraping

Slack Notification Agent

🟢 Safe

Watch for events (deploys, errors, mentions) across your stack and route smart notifications to the right Slack channel with context.

DevOps alerts, team coordination, incident response

Capability Stack

Slack SendComposio🟢
Slack MembersComposio🟢
GitHub EventsComposio🟢
pythonSDK docs →
from agentsift import AgentSift

client = AgentSift()

# All safe Slack capabilities
slack_caps = client.capabilities.list(
    platform="composio",
    q="slack",
    risk="safe",
)
for cap in slack_caps:
    print(f"{cap.name}: trust={cap.trust_score:.2f}")
#slack#notifications#devops

Data Pipeline Agent

🟡 Caution

Query databases, transform data, and write results to storage or downstream APIs. The agent-native ETL.

Data engineering, reporting, analytics

Capability Stack

PostgreSQL QueryComposio🟡
Write FileMCP🟡
Drive UploadComposio🟡
typescriptSDK docs →
import { AgentSift } from 'agentsift';

const client = new AgentSift();

// Compare DB access vs file write risk
const comparison = await client.compare(
  'composio-postgresql-run-postgresql-query',
  'mcp-filesystem-write-file',
  'composio-googledrive-upload-file',
);
// Shows shared permissions, high-risk perms, verdict
console.log(comparison.verdict);
console.log('High-risk perms:', comparison.permissionMatrix.highRisk);
#database#etl#data

Missing a recipe? The full capability registry has 6,000+ tools to build from.

Search CapabilitiesAPI Docs