VERO Engineering

OpenCode

AI-Powered Development & DevOps Automation

opencode

you:

Navigate with Space or Arrow Keys

The Traditional Way

Debugging a production incident manually...

Manual Process (15-30 min)

# 1. Check Slack for context - Open browser, find channel - Scroll through messages - Copy error details # 2. Check GitLab for recent deploys - Open GitLab UI - Navigate to pipelines - Check merge requests # 3. Search logs - Open Grafana/Kibana - Write query manually # 4. Check AWS/K8s state - aws cli commands - kubectl commands # 5. Report back to Slack

With OpenCode (2-3 min)

you: Check the payment timeout issue in #incidents and debug it # OpenCode automatically: ✓ Reads Slack thread context ✓ Queries GitLab for recent changes ✓ Searches logs with smart filters ✓ Checks cluster health via AWS ✓ Correlates all findings ✓ Posts summary to Slack thread All in one natural language request

The OpenCode Philosophy

OpenCode doesn't just run scripts.
It orchestrates your entire toolkit through natural conversation.

💬

Natural
Language

→

🧠

AI
Planning

→

⚡

Skill
Execution

→

🔄

Feedback
Loop

DEVELOPMENT WORKFLOWS

Skills While Coding

OpenCode isn't just for DevOps - it supercharges your development workflow

🔍 Context Gathering

Pull Jira ticket details, Confluence docs, and Slack discussions into your coding session

📊 Research & Design

Use deep-research.js for architecture decisions with multi-source analysis

🔗 Cross-Repo Edits

Make coordinated changes across client, API, and backend services

📝 Documentation

Auto-update Confluence, post to Slack, and sync Jira status

Press ↓ to see examples

Example: Start a Feature with Full Context

opencode — feature development

you: I'm starting VSB-1456, get me the full context

◐ Fetching context from multiple sources...

→ jira-search.js "key = VSB-1456"

→ confluence-search.js "VSB-1456 OR user-preferences"

→ slack-unreads.js --content --channels="backend-dev"

Context assembled:

Jira Ticket: Add user preference sync across devices

Acceptance: Real-time sync, offline support, conflict resolution

Related: VSB-1420 (auth refactor), VSB-1389 (storage layer)

Confluence: "User Preferences Architecture" doc found

Shows: Event-driven sync via Kafka, Redis cache layer

Slack: @backend mentioned conflict resolution approach yesterday

Decision: Last-write-wins with timestamp, UI shows conflicts

Example: Architecture Research

opencode — architecture design

you: research how we should implement rate limiting for the API gateway

◐ Starting deep research...

→ deep-research.js "API rate limiting strategies"

→ confluence-search.js "rate limit OR throttling"

→ gitlab-search "rate.limit" --scope=code

Research complete. Summary:

Existing Implementation:

• auth-service: Token bucket (100 req/min per user)

• payment-service: Fixed window (10 req/sec)

Recommendation for API Gateway:

• Sliding window counter (Redis ZSET)

• Consistent with AWS API Gateway patterns

CROSS-REPO EDITING

Multi-Repository Feature Implementation

Coordinated changes across your entire stack

web-client

React + TypeScript

→

api-gateway

Node.js + Express

→

user-service

Go + gRPC

→

notification-svc

Python + Kafka

Scenario: Add "Email Notification Preferences" Feature

UI toggle in settings page (web-client)
New API endpoint (api-gateway)
Preference storage (user-service)
Preference check before sending (notification-svc)

Press ↓ to see the workflow

Step 1: Understand the Stack

opencode — cross-repo feature

you: I need to add email notification preferences. Show me how

preferences flow through our stack currently.

◐ Analyzing codebase across repos...

→ Searching web-client/src for preference patterns

→ Searching api-gateway/routes for /preferences

→ Searching user-service for preference storage

Current preference flow mapped:

web-client/

src/components/Settings/PreferencesPanel.tsx

src/api/userPreferences.ts → calls /api/v1/preferences

api-gateway/

routes/preferences.js → proxies to user-service

user-service/

internal/preferences/handler.go → stores in PostgreSQL

internal/preferences/model.go → Preferences struct

Step 2: Modify Backend Services

opencode — backend changes

you: add email_notifications boolean to the preferences model

and update the handler to accept it

◐ Modifying user-service...

✓ Edited user-service/internal/preferences/model.go

+ EmailNotifications bool `json:"email_notifications" db:"email_notifications"`

✓ Edited user-service/internal/preferences/handler.go

+ Added field to UpdatePreferences request

+ Added validation (defaults to true for new users)

✓ Created user-service/migrations/003_add_email_prefs.sql

ALTER TABLE preferences ADD COLUMN email_notifications BOOLEAN DEFAULT true;

◐ Running tests...

✓ All 47 tests passing

Step 3: Update API Gateway

opencode — api gateway

you: update the api-gateway to pass through the new field

and add OpenAPI documentation

◐ Updating api-gateway...

✓ Edited api-gateway/routes/preferences.js

+ email_notifications in request validation schema

✓ Edited api-gateway/openapi/preferences.yaml

+ Added email_notifications property (boolean)

+ Added description and default value

✓ Edited api-gateway/tests/preferences.test.js

+ Test case for email_notifications toggle

◐ Running tests...

✓ All 23 tests passing

Step 4: Update Web Client

opencode — frontend

you: add a toggle for email notifications in the preferences panel

match the existing style and add proper types

◐ Updating web-client...

✓ Edited web-client/src/types/preferences.ts

+ emailNotifications: boolean

✓ Edited web-client/src/api/userPreferences.ts

+ Updated API calls to include new field

✓ Edited web-client/src/components/Settings/PreferencesPanel.tsx

+ Added Toggle component with label

+ Connected to preferences state

+ Added loading state during save

◐ Running tests and type check...

✓ Types: No errors

✓ Tests: 156 passing

Step 5: Update Notification Service

opencode — notification service

you: update notification-svc to check user preferences

before sending email notifications

◐ Updating notification-svc...

✓ Edited notification-svc/src/handlers/email_handler.py

+ Added preferences check before send_email()

+ Calls user-service gRPC to fetch preferences

✓ Edited notification-svc/src/clients/user_client.py

+ Added get_email_preference() method

+ Includes caching (5 min TTL)

✓ Edited notification-svc/tests/test_email_handler.py

+ Test: skip when preference disabled

+ Test: send when preference enabled

◐ Running pytest...

✓ 34 tests passing

Cross-Repo Edit Summary

web-client

3 files changed

→

api-gateway

3 files changed

→

user-service

3 files changed

→

notification-svc

3 files changed

Files Modified

Languages

260

Tests Passing

~10 min

Total Time

One conversation. Full stack feature. All tests passing.

INCIDENT RESPONSE DEMO

Use Case: Production Incident

A timeout alert fires. You need to investigate.

Press ↓ to see the workflow

Step 1: Gather Context from Slack

opencode — incident response

you: what's happening in #incidents about payment timeouts?

◐ Reading Slack channel #incidents...

→ slack-unreads.js --channels=incidents --content

Found 3 relevant messages:

@oncall: Payment service timing out since 14:32 UTC

@backend: Seeing 502s on /api/payments/process

@platform: Might be related to yesterday's deploy?

Step 2: Check Recent Deployments

opencode — incident response

opencode: Checking GitLab for recent payment service changes...

→ gitlab-activity-flexible.js 2024-11-28..2024-11-30

Recent deployments to payment-service:

⚠ MR !2847 "Add retry logic for upstream calls"

Merged: yesterday 16:45 UTC by @backend-dev

Changed: src/services/upstream.js (+47, -12)

○ MR !2841 "Update dependencies" - 2 days ago

Likely culprit: MR !2847 introduced new retry logic

Step 3: Search Logs for Evidence

opencode — incident response

opencode: Searching logs for timeout patterns...

→ vero-logs-search.js --service=payment-service --level ERROR --hours 2

Found 847 errors in last 2 hours (normal baseline: ~20)

Top error patterns:

[623x] TimeoutError: upstream request exceeded 30s

[198x] RetryExhausted: max retries (5) reached

[ 26x] ConnectionReset: upstream closed connection

Pattern: New retry logic is amplifying upstream latency

Step 4: Post Summary to Slack

opencode — incident response

you: post the findings to the incident thread

→ slackFetch('chat.postMessage', { channel: 'incidents', thread_ts: ... })

✓ Posted to #incidents thread:

*Investigation Summary* AI generated Root Cause: Cascade failure from auth-service degradation • auth-service p99 latency spiked to 4.2s (normally 200ms) • MR !2847 (yesterday) added retry logic: 5 retries × 30s timeout • Combined effect: 150s worst-case per request Immediate Fix: Reduce retry count to 2, add circuit breaker Long-term: Investigate auth-service memory pressure

Full Incident Workflow

💬

Slack

Read context

🦊

GitLab

Find changes

📊

Loki/Logs

Search errors

☁️

AWS/K8s

Check state

✅

Slack

Report back

Total time: ~2 minutes vs 15-30 minutes manually

Use Case: Automated Standup

Generate your daily update from actual work data

Press ↓ to see it in action

opencode — daily standup

you: generate my standup for today

◐ Gathering activity from multiple sources...

→ gitlab-personal-activity.js yesterday

→ jira-search.js "assignee = currentUser() AND updated >= -1d"

→ slack-unreads.js --content --channels="backend-dev"

Generated standup:

**Yesterday:** • Merged MR !2851: API rate limiting implementation • Code review on !2849 (auth refactor) - requested changes • VSB-1234: Moved to "In Review" after addressing feedback **Today:** • VSB-1238: Start database migration planning • Follow up on !2849 review comments **Blockers:** • Waiting on DevOps for staging environment access

Use Case: MR Review Analysis

AI-powered analysis of review comments with cross-cutting insights

Press ↓ to explore

opencode — code review analysis

you: analyze the review comments on MR !2847

→ gitlab-mr-review-comments-analyzer.js https://gitlab.../merge_requests/2847

◐ Fetching Jira context for VSB-1234...

◐ Fetching GitLab code context...

Analysis of 12 review comments:

Cross-cutting patterns identified:

1. Error handling (3 comments) - missing try/catch blocks

2. Logging concerns (2 comments) - debug logs in prod code

3. Type safety (2 comments) - any types in interfaces

Priority breakdown:

🔴 CRITICAL: 1 🟠 HIGH: 3 🟡 MEDIUM: 5 🟢 LOW: 3

Full report with AI-generated responses available for each comment.

BROWSER AUTOMATION

Use Case: Browser Control

Automate web interactions with AI-powered element detection

🎯 AI-Powered Clicking

smartClick('login button') — AI finds and clicks elements by natural language

📸 Set-of-Marks (SoM)

Industry-standard visual grounding — annotates screenshots for precise element selection

🔐 Profile & Cookie Support

Use Zen/Firefox profiles with saved logins — no need to re-authenticate

🍪 Auto Overlay Dismissal

Automatically handles cookie banners, consent dialogs, and modal popups

Press ↓ to see examples

Example: Scrape Internal Dashboard

opencode — browser automation

you: export the monthly analytics from the admin dashboard

◐ Starting browser session with Zen profile...

→ createSession('zen', { headless: false })

✓ Session started (using release profile with saved cookies)

→ navigate('https://admin.company.com/dashboard')

🍪 Dismissed 2 overlay(s): cookie-banner, modal

✓ Page loaded — already authenticated via profile

→ smartClick('Analytics tab')

🤖 AI selected [4]: nav link with "Analytics" text (0.95)

→ smartClick('Export CSV')

📜 Scrolling element into view...

👆 Clicking at viewport position (847, 412)

✓ Download triggered: analytics-nov-2024.csv

Example: Fill and Submit Web Form

opencode — form automation

you: create a support ticket for the payment API timeout issue

◐ Opening support portal...

✓ Navigated to support.company.com/new-ticket

→ smartClick('Category dropdown')

→ smartClick('API Issues option')

→ smartClick('Subject input')

→ humanType('Payment API timeout errors')

→ smartClick('Description textarea')

→ type('Seeing 30s timeouts on /api/payments/process...')

→ smartClick('Priority dropdown')

→ smartClick('High priority')

→ submitForm()

✓ Ticket created! ID: SUPPORT-4821

The API

browser-control.js

// Create persistent session

const { createSession } = await import('~/.../browser-control.js');

const s = await createSession('zen', { headless: false });

// AI-powered interaction (auto-scrolls if needed)

await s.navigate('https://example.com'); // auto-dismisses overlays

await s.smartClick('search box'); // AI finds + scrolls + clicks

await s.humanType('my query'); // types with human-like delays

await s.press('Enter'); // press any key

await s.fullPageScreenshot('/tmp/full.png'); // entire page

await s.close();

Browser Engine Options

Playwright (chrome/chromium/edge) — Fast, stealth, CAPTCHA detection
Selenium (zen/firefox/safari) — Real profiles with saved sessions

Key Session Methods

Navigation & Interaction

navigate(url) — auto-dismisses overlays
smartClick(description) — AI-powered click
clickAt(x, y) — coordinate click
type(text) / humanType(text)
press(key) — Enter, Tab, Escape...
submitForm() — submit active form

Page Analysis

getAllClickableElements() — entire page
getClickableElements() — viewport only
getContext() — page text & metadata
getPageInfo() — dimensions & scroll
detectCaptcha() — check for blocks
getCookies() — read all cookies

Screenshots & Scrolling

screenshot(path) — viewport capture
fullPageScreenshot(path) — entire page
scroll('down', 400) — by direction
scrollTo(x, y) — absolute position
scrollElementIntoView(el) — center element

Advanced Features

dismissOverlays() — banners/modals
solveCaptcha() — 2captcha or manual
clickByDescriptionSoM() — Set-of-Marks
evaluate(js) — run JS in page
waitFor(selector) — wait for element

Browser Automation Flow

🚀

Create Session

Load profile

🌐

Navigate

Dismiss overlays

🔍

Find Elements

AI analysis

📜

Auto-Scroll

Into viewport

👆

Click/Type

At coordinates

smartClick() combines AI element detection + auto-scroll + coordinate clicking — one method does it all

DATABASE OPERATIONS

Use Case: Cassandra + Schema Knowledge Base

Query clusters across environments with AI-powered schema understanding

🧠 Schema KB

SQLite knowledge base with 86 tables, relationships, and usage patterns — AI-queryable

🔗 Relationships

Understand how tables connect: feeds_into, partitioned_by, aggregates, filters

🔐 Automatic Auth

JumpCloud → Vault auth, SSH tunnels for prd — no prompts needed

⚡ Daemon Mode

Persistent connections via background daemon — ~15ms query latency

Press ↓ to see examples

Example: Query User Data Across Environments

opencode — cassandra query

you: check if user abc123 exists in vero.users on dev

◐ Connecting to Cassandra...

→ cassandra-query.js dev cass000 "SELECT * FROM vero.users WHERE user_id = 'abc123'"

🔐 Authenticating to Vault via JumpCloud LDAP...

Query returned 1 row:

user_id: abc123

email: user@example.com

created_at: 2024-03-15 10:30:00

status: active

you: now check the same user in prd

→ cassandra-query.js prd cass000 "SELECT * FROM vero.users WHERE user_id = 'abc123'"

🔒 Establishing SSH tunnel through bastion...

Query returned 1 row (prd):

user_id: abc123

status: suspended

Note: User status differs between dev (active) and prd (suspended)

Example: Auto-Discover Table Location

opencode — table discovery

you: which cluster has the payment_events table?

→ cassandra-cache-mappings.js --find payment.payment_events

Found payment.payment_events:

dev: cass002

stg: cass002

prd: cass002

you: show me the schema for that table

→ cassandra-query.js dev cass002 "DESCRIBE TABLE payment.payment_events"

Table schema:

CREATE TABLE payment.payment_events (

event_id uuid,

user_id text,

event_type text,

amount decimal,

created_at timestamp,

PRIMARY KEY ((user_id), created_at, event_id)

) WITH CLUSTERING ORDER BY (created_at DESC);

Example: Daemon Mode for Fast Queries

opencode — cassandra daemon

you: I need to run several queries on dev cass000, use daemon mode

→ cassandra-daemon.js start dev cass000

Starting cassandra daemon for dev/cass000...

✓ Daemon started and ready (socket: /tmp/cassandra-daemon-dev-cass000.sock)

you: query 1: count users created today

→ cassandra-query.js dev cass000 --daemon "SELECT COUNT(*) FROM vero.users..."

✓ 847 users (query: 23ms)

you: query 2: check failed payments

→ cassandra-query.js dev cass000 --daemon "SELECT * FROM payment.events..."

✓ 12 failed payments (query: 18ms)

you: query 3: get session counts

→ cassandra-query.js dev cass000 --daemon "SELECT COUNT(*) FROM auth.sessions..."

✓ 15,432 sessions (query: 15ms)

Daemon auto-shuts down after 30min idle

Search Tables by Pattern

opencode — table search

you: find all tables related to streaming

→ cassandra-cache-mappings.js --search "*stream*"

Found 8 matching tables:

streaming.events dev:cass001 stg:cass001 prd:cass001

streaming.sessions dev:cass001 stg:cass001 prd:cass001

streaming.analytics dev:cass001 stg:cass001 prd:cass001

vero.stream_history dev:cass000 stg:cass000 prd:cass000

vero.stream_metadata dev:cass000 stg:cass000 prd:cass000

analytics.stream_events dev:cass002 stg:cass002 prd:cass002

analytics.stream_metrics dev:cass002 stg:cass002 prd:cass002

logs.stream_errors dev:cass003 stg:cass003 prd:cass003

you: list all tables in the vero keyspace

→ cassandra-cache-mappings.js --list vero

vero.users, vero.sessions, vero.profiles, vero.stream_history...

Total: 24 tables in vero keyspace

Schema Knowledge Base: Semantic Search

opencode — schema knowledge base

you: where are user feeds stored?

→ schema-kb.js --search "user feed timeline"

Found in Schema KB:

vero_social_stream.streams

Main user feed/timeline storage. Posts from followed

users are fan-out written here for efficient retrieval.

you: how do posts get to user feeds?

→ schema-kb.js --relations vero_social_post.posts

Relationships:

feeds_into → vero_social_stream.streams

Posts are distributed to user streams/feeds

aggregates → vero_social_post_counter.likes

aggregates → vero_social_post_counter.comments

Schema Knowledge Base: Full Table Details

opencode — table details

you: show me full details for the streams table

→ schema-kb.js --table vero_social_stream.streams

## vero_social_stream.streams

Domain: feeds

Description: Main user feed/timeline storage

### Usage

StreamsRepository (read, write)

→ social/domain/StreamsRepository.java

### Relates To

partitioned_by → vero_social_stream.epochs

### Referenced By

vero_social_post.posts feeds_into →

vero_social_follow.user_followers determines_content →

vero_social_hidden_stream.hidden_stream filters →

Cassandra + Schema KB Overview

🧠

schema-kb

Understand schema

🔍

cache-mappings

Find clusters

⚡

daemon

Fast queries

🔐

Vault

Auto auth

📊

query

Execute CQL

Tables in KB

Relationships

271

Query Patterns

~15ms

Daemon Queries

Ask: "where are user feeds stored?" → Schema KB finds the table → OpenCode queries it

Use Case: Cross-Platform Search

Find information wherever it lives

opencode — unified search

you: find everything about the OAuth migration project

◐ Searching across platforms in parallel...

→ confluence-search.js "OAuth migration"

→ jira-search.js "text ~ 'OAuth migration'"

→ gitlabFetch('/api/v4/search?scope=merge_requests&search=oauth')

→ slackFetch('search.messages', { query: "OAuth migration" })

Found 23 results:

📄 Confluence: "OAuth 2.0 Migration Plan" (updated 3 days ago)

🎫 Jira: VSB-892 "Implement OAuth2 flow" (In Progress)

🦊 GitLab: MR !2834 "OAuth provider refactor" (Open)

💬 Slack: #backend discussion thread (yesterday)

How It Works

Skills

Modular JavaScript/TypeScript scripts for each integration

Identity

Secure credential management via ~/.secrets/

Runtime

Bun for fast execution with native TypeScript support

Orchestration

AI plans and chains skills based on natural language intent

Caching

Smart caching of users, channels, and frequently accessed data

Audio Feedback

TTS notifications via Kokoro for long-running operations

skill execution

$ bun ~/.config/opencode/skills/<category>/<skill>.js [args]

Integration Ecosystem

💬

Slack

🦊

GitLab

🎫

Jira

📚

Confluence

☁️

AWS

📊

Loki/Grafana

🗄️

Cassandra

🌐

Browser

🎮

Discord

🎵

Spotify

📱

macOS

🤖

AI/LLMs

AUDIO FEEDBACK

Text-to-Speech Notifications

Kokoro TTS provides audio feedback for long-running operations

Automatic Policy

TTS triggers automatically after 3+ operations complete — keeps you informed while multitasking

Multi-Voice Support

40+ voices including US/UK English, French, Italian, Japanese, and Mandarin Chinese

Background Execution

Runs async (&) so it doesn't block the AI — audio plays while work continues

Local & Fast

Kokoro runs locally — no API calls, instant response, works offline after model download

Press down to see usage

TTS Usage Examples

tts-kokoro.js

// Policy from AGENTS.md - auto TTS after 3+ ops

bun -e "const{speak}=await import(

'$HOME/.config/opencode/skills/media/tts-kokoro.js');

await speak('Done');" &

// CLI usage

$ bun ~/.../tts-kokoro.js "Task complete"

$ bun ~/.../tts-kokoro.js "Bonjour" --voice ff_siwis

$ bun ~/.../tts-kokoro.js "Fast" --speed 1.5

// Programmatic API

const { speak } = await import('~/.../tts-kokoro.js');

await speak('Build succeeded', { voice: 'am_adam' });

Available Voice Categories

US English: af_sarah, am_adam, af_nova, af_sky, am_michael, af_bella, am_fenrir...
British: bf_alice, bf_emma, bf_isabella, bm_george, bm_lewis, bm_daniel...
French: ff_siwis
Japanese: jf_alpha, jf_gongitsune, jm_kumo, jf_nezumi, jf_tebukuro...
Chinese: zf_xiaobei, zf_xiaoni, zf_xiaoxiao, zf_xiaoyi, zm_yunjian, zm_yunxi...
Italian: if_sara, im_nicola
Portuguese: pf_dora, pm_alex, pm_santa
Hindi: hf_alpha, hf_beta, hm_omega, hm_psi

Configuration Architecture

Five key files power the OpenCode skills ecosystem

📋

AGENTS.md

AI Instructions

⚙

opencode.json

App Config

🚀

skill-init.js

Environment Setup

👤

identity.env

Personal Info

🔒

vdp.env

Infra Secrets

Press down to explore each file

AGENTS.md — AI System Instructions

~/.config/opencode/AGENTS.md

# OpenCode Skills

**Runtime**: `bun` | **Secrets**: `~/.secrets/`

**Identity**: `~/.config/opencode/identity.env`

## Policies

- **TTS** (3+ ops): `bun -e "...speak('Done');" &`

- **Parallel**: Multiple Task tools in ONE response

- **AI disclaimer**: Slack: `*AI generated*`

## Skills

| workflows | daily-standup.js | standup |

Purpose

Injected into AI context — AI reads on every session
Skill catalog — all skills with triggers
Policies — TTS, parallel execution
URLs — internal service endpoints

Key Sections

Skill table with triggers and examples
Security policies (never print keys)
Browser automation API reference

opencode.json — Application Configuration

~/.config/opencode/opencode.json

{

"$schema": "https://opencode.ai/config.json",

"theme": "opencode",

"model": "openrouter/anthropic/claude-sonnet-4.5",

"small_model": "openrouter/google/gemini-2.5-flash-lite",

"provider": {

"openrouter": {

"options": {

"apiKey": "{file:../../.secrets/openrouter-key}"

}

"autoupdate": true

}

Purpose

Model selection — which AI models to use
Provider config — API keys, endpoints
Thinking mode — extended reasoning tokens
MCP servers — Model Context Protocol

Key Features

{file:...} loads secrets from files
JSON Schema validation
Multi-provider support

skill-init.js — Environment Bootstrap

bun ~/.config/opencode/skill-init.js

Checking System Dependencies

+ Bun JavaScript runtime (bun)

+ Git version control (git)

+ Kokoro TTS (kokoro-tts)

Checking Identity Configuration

+ Identity file (identity.env)

Checking API Keys

+ Atlassian (~/.secrets/atlassian-key)

+ Slack (~/.secrets/slack-xoxp-key)

+ GitLab (~/.secrets/gitlab-key)

Initializing Caches

+ Jira users cache

+ Slack channels cache

Environment fully configured!

Purpose

Environment validation — checks all deps
API key verification — validates secrets
Auto-setup — creates identity.env from template
Security enforcement — fixes file permissions

What It Checks

System deps: bun, git, curl, python, kokoro-tts
API keys: Atlassian, Slack, GitLab, OpenRouter
Directories: ~/.secrets/, ~/.config/opencode/
Caches: Jira/Slack/GitLab users, channels
Installs: browser automation drivers

Run with --check-only to verify without installing, or --force to refresh all caches

identity.env — Personal Information

~/.config/opencode/identity.env

# Personal identity for AI context

FULL_NAME="Your Full Name"

EMAIL="you@company.com"

SLACK_USER_ID="U12345678"

GITLAB_USERNAME="yourusername"

JIRA_USERNAME="you@company.com"

TIMEZONE="America/New_York"

LOCATION="City, Country"

Purpose

Personalization — AI knows who you are
Cross-platform identity — maps accounts
Automatic filtering — "my commits" works

How It's Used

GitLab: Filter personal activity reports
Jira: Default assignee for searches
Slack: Identify your user ID
Standups: "What did I do yesterday?"

Auto-created from identity.env.template on first run of skill-init.js

vdp.env — Infrastructure Secrets

~/.config/opencode/vdp.env

# Infrastructure credentials (chmod 0600)

VAULT_ADDR_TEMPLATE="https://vault.{env}.vdp.vero.host"

JUMPCLOUD_USERNAME="your_username"

JUMPCLOUD_PASSWORD="your_password"

# Grafana hosts for log searches

VDP_DEV_HOST="grafana.dev.vdp.vero.host"

VDP_STG_HOST="grafana.stg.vdp.vero.host"

VDP_PRD_HOST="grafana.prd.vdp.vero.host"

Purpose

Vault auth — JumpCloud LDAP credentials
Non-interactive — No password prompts
Token caching — Auto-renewed per env

Used By Skills

cassandra-query.js — Vault → DB creds
cassandra-daemon.js — Persistent conns
vero-logs-search.js — Loki logs

Security: Never print these values. JUMPCLOUD_PASSWORD enables fully automatic Vault auth (no prompts).

Pre-Built Workflows

End-to-end automation in ~/.config/opencode/skills/workflows/

🌅 daily-standup.js

GitLab commits + Jira updates + Slack + Calendar into one standup report

📊 activity-report.js

Team activity from Slack + GitLab with TODOs and blockers

🔍 jira-review-mr-workflow.js

Fetch review tickets, analyze MRs, post AI reviews to GitLab + Jira

🦊 post-gitlab-activity.js

GitLab activity report posted to Slack with threaded sections

📝 opencode-commit-publish.js

Commit + push + AI release notes to Slack in one command

🔧 Build Your Own

Compose skills into custom workflows with shared APIs

All support --dry-run --verbose --output — or invoke via natural language

NOTABLE NEW FEATURES

Recent Skill Enhancements

New capabilities across the skills ecosystem

🧠 Knowledge Bases

SQLite-backed KBs for Schema, Git Repos, and VDP Gateway API — semantic search & relationships

💬 Enhanced Slack

Mark channels as read, upload files, date range filtering, wildcard channel matching

📱 macOS Integrations

WhatsApp send via Baileys, Calendar CRUD, Mail operations, Messages access

🎮 Discord & AI

AI summaries, attachment downloads, deep research with web search & extended thinking

Press ↓ to explore each category

Knowledge Base Skills

🗄️ schema-kb.js

Cassandra schema with semantic search

86 tables with descriptions
Relationship mapping
--search "user feed"
--relations streams

📁 git-repos-kb.js

Local git repository indexing

Full-text search
Language & framework filter
--lang TypeScript
--commits repo 20

🔌 vdp-gateway-kb.js

OpenAPI spec discovery

Endpoint & schema explore
Auto-refresh from remote
--endpoint /api/x GET
--tag authentication

knowledge base queries

you: which tables store user timelines?

→ schema-kb.js --search "user timeline feed"

Found: vero_social_stream.streams — Main feed storage

Relationships: posts feeds_into → streams

Enhanced Slack Skills

slack-mark-read.js

# Mark all channels as read

$ bun slack-mark-read.js --all

✓ Marked 12 channels as read

# Mark specific channel

$ bun slack-mark-read.js backend-dev

✓ Marked #backend-dev as read

slack-upload-file.js

# Upload file to channel

$ bun slack-upload-file.js \

--file=/tmp/report.pdf \

--target=backend-dev --title="Report"

✓ Uploaded to #backend-dev

Other Slack Improvements

--oldest / --latest date filtering
--channels="backend-*" wildcards
--include-read fetch read messages
--with-directives AI summary hints
--thread-ts for thread replies
DM and group message support

macOS System Integrations

📅 Calendar

List/search events
Create events
Update & delete
Multi-calendar

📧 Mail

List inbox
Read messages
Download attachments
Archive/delete/send

💬 WhatsApp

Send text messages
Send images
QR auth (Baileys)
List contacts

💬 Messages

Read iMessage/SMS
Search conversations
History access
Read-only

macos skills examples

you: what meetings do I have tomorrow?

→ macos-calendar.js list 2025-12-03

Found 3 events: 09:00 Standup, 14:00 Design Review, 16:30 1:1

you: send WhatsApp to John: "Running 10 min late"

→ macos-whatsapp-send.js --send --to "+1..." --message "..."

✓ Message sent to John

Discord & AI Enhancements

discord-fetch-messages.js

# Fetch with AI summary

$ bun discord-fetch-messages.js --channel="123" --ai-summary

# Download attachments

$ bun discord-fetch-messages.js --channel="123" --download-attachments

# By server & channel name

$ bun discord-fetch-messages.js --server="OpenCode" --channel-name="general"

deep-research.js

# Opus 4.5 + Exa web search

$ bun deep-research.js "GraphQL federation patterns"

# Control reasoning depth

$ bun deep-research.js --reasoning-effort=high "Design ML pipeline"

# Pure reasoning (no web)

$ bun deep-research.js --no-search "Analyze algorithm"

Deep Research Features

Claude Opus 4.5 extended thinking
Exa web search integration
Streaming with visible reasoning
Configurable reasoning effort
Up to 20 search results
System message customization

Confluence & Jira Improvements

confluence-post-comment.js

Post comments to any page
--reply to reply to comments
--ai adds AI generated marker
Supports stdin for long content
Accepts page ID or full URL

jira-search.js

Full JQL support with examples
--comments include comments
--expand changelog for history
--fetch-all pagination
--json structured output

jira & confluence

you: find all my open tickets with comments

→ jira-search.js "assignee = currentUser() AND resolution is EMPTY" --comments

Found 8 issues with comments

you: add a note to the VSB-1234 Confluence page

→ confluence-post-comment.js 4067262513 --ai '<p>Analysis...</p>'

✓ Comment posted (AI generated)

Transform Your Workflow

10x

Faster debugging

Context switches

∞

Composability

One interface. All your tools. Natural language.

🔊 Audio feedback when tasks complete | 🔒 Secure secrets management | 👤 Personal identity context

Get Started

Start Automating

get started

$ bun ~/.config/opencode/skill-init.js

# Validates deps, keys, identity, and initializes caches

Just describe what you need. OpenCode handles the rest.