slack-agent

Slack Agent Development Skill

This skill builds Slack agents with eve — Vercel's filesystem-first framework for durable backend agents — using Vercel Connect for Slack credentials:

• eve (eve package) — agent runtime, tools, channels, durability
• @vercel/connect — brokered short-lived Slack tokens; no bot tokens or signing secrets to manage

Skill Invocation Handling

When this skill is invoked via /slack-agent, check for arguments and route accordingly:

Command Arguments

Auto-Detection (No Argument)

If invoked without arguments, detect the project state and route appropriately:

1. No package.json with eve and no agent/ directory → Treat as new, start Phase 1
2. Has eve project but no agent/channels/slack.ts → Start Phase 2 (Slack connector + channel)
3. Has Slack channel but no SLACK_CONNECTOR configured → Start Phase 3
4. Configured but not deployed → Start Phase 5 (the Slack surface only works on a deployment)
5. Deployed but no tests → Start Phase 6
6. Otherwise → Provide general assistance using this skill's patterns

Project Detection

Detect an eve project by either signal:

• package.json contains "eve" as a dependency
• An agent/ directory with instructions.md and/or agent.ts exists

If neither is present, this is a new project: scaffold with npx eve@latest init (Node 24+ required).

Wizard Phases

The wizard is located in ./wizard/ with these phases:

• 1-project-setup.md - Understand purpose, generate custom implementation plan, scaffold with npx eve@latest init
• 1b-approve-plan.md - Present plan for user approval before scaffolding
• 2-create-slack-app.md - Create the Slack connector with Vercel Connect and add the Slack channel
• 3-configure-environment.md - Set up env vars (SLACK_CONNECTOR, model credentials)
• 4-test-locally.md - Test agent logic locally with the eve dev TUI (Slack surface tests happen after deploy)
• 5-deploy-production.md - Deploy with eve deploy, verify the Slack surface
• 6-setup-testing.md - Vitest configuration

IMPORTANT: For new projects, you MUST:

1. Read ./wizard/1-project-setup.md first
2. Ask the user what kind of agent they want to build
3. Generate a custom implementation plan using ./reference/agent-archetypes.md
4. Present the plan for approval (Phase 1b) BEFORE scaffolding the project
5. Only proceed to scaffold after the plan is approved

General Development Guidance

You are working on a Slack agent project built with eve. Follow these mandatory practices for all code changes.

Project Stack

• Framework: eve (filesystem-first agent framework; Node 24+)
• Slack channel: eve/channels/slack + @vercel/connect for credentials
• AI: model routed through Vercel AI Gateway by default (anthropic/claude-sonnet-5); tool schemas with zod
• Durability: Workflow SDK under the hood (Vercel Workflows when deployed on Vercel)
• Linting: Biome
• Package Manager: pnpm (or npm — npx eve@latest init installs with npm)

{
  "engines": { "node": "24.x" },
  "dependencies": {
    "eve": "latest",
    "ai": "latest",
    "zod": "^3.x",
    "@vercel/connect": 

Filesystem-First Layout

In eve, a file's location says what it does, and its path usually gives it its name. The whole agent lives under agent/:

Even a two-file agent (instructions.md + agent.ts) gets file, shell, web, and delegation tools out of the box from the default harness. Full docs are bundled at node_modules/eve/docs/ once eve is installed — read them when a detail isn't covered here.

Quality Standards (MANDATORY)

These quality requirements MUST be followed for every code change. There are no exceptions.

After EVERY File Modification

1. Run linting immediately:

   pnpm lint
   • If errors exist, run pnpm lint --write for auto-fixes
   • Manually fix remaining issues
   • Re-run pnpm lint to verify

2. Check for corresponding test file:

   • If you modified foo.ts, check if foo.test.ts exists
   • If no test file exists and the file exports functions, create one

Before Completing ANY Task

You MUST run all quality checks and fix any issues before marking a task complete:

# 1. TypeScript compilation - must pass
pnpm typecheck
 
# 2. Linting - must pass with no errors
pnpm lint
 
# 3. Tests - all tests must pass
pnpm test

Do NOT complete a task if any of these fail. Fix the issues first.

Unit Tests Required

For ANY code change, you MUST write or update unit tests.

• Location: Co-located *.test.ts files (e.g. agent/tools/get_weather.test.ts)
• Framework: Vitest
• Coverage: All exported functions and every tool's execute() (including error paths) must have tests

Example test structure:

import { describe, it, expect, vi } from 'vitest';
import { myFunction } from './my-module';
 
describe('myFunction', () => {
  it('should handle normal input', () => {
    expect(myFunction('input')).toBe('expected');
  });
 
  it('should handle edge cases', () => {
    expect(myFunction('')).toBe('default');
  });
});

E2E Tests for User-Facing Changes

If you modify:

• Dispatch hooks (onAppMention, onDirectMessage, onInteraction)
• Custom channel event handlers
• Tools the agent calls in response to Slack messages
• Delivery behavior (what gets posted to Slack)

You MUST add or update tests that verify the full flow. Remember: the Slack surface itself cannot be exercised locally (see Gotchas), so E2E coverage means unit/integration tests around your handlers plus a post-deploy smoke test.

Bot Setup Patterns (CRITICAL)

Slack Channel (agent/channels/slack.ts)

The Slack channel is a single file. Its filename registers the slack channel, served at /eve/v1/slack — this is the canonical trigger path everywhere in this skill.

// agent/channels/slack.ts
import { connectSlackCredentials } from "@vercel/connect/eve";
import { slackChannel } from "eve/channels/slack";
 
export default slackChannel({
  credentials: connectSlackCredentials(process.env.SLACK_CONNECTOR!),
});

connectSlackCredentials(connectorUid) returns { botToken, webhookVerifier }:

• botToken — resolved at runtime via Vercel Connect as a short-lived, app-scoped token; Connect handles rotation and multi-workspace tenancy
• webhookVerifier — confirms each forwarded event genuinely came from Connect (replaces Slack's native signature check)

There is no SLACK_BOT_TOKEN and no SLACK_SIGNING_SECRET in this stack. The only Slack env var is SLACK_CONNECTOR (the connector UID, e.g. slack/my-agent).

Vercel Connect Setup

Create a Slack connector and point its trigger at eve's Slack route:

npm install -g vercel@latest
 
# Create the connector with event forwarding enabled
vercel connect create slack --triggers
 
# Attach the project as a trigger destination on eve's route
# (the default trigger path is /slack — set it explicitly):
vercel connect attach <uid> --triggers --trigger-path /eve/v1/slack --yes

--triggers is required. Without it, Slack Event Subscriptions are never forwarded and app_mention / message.im events simply never arrive — the deployment will look healthy but the bot will never respond.

You can also add the channel with eve channels add slack, which scaffolds agent/channels/slack.ts for you.

Deploy

eve deploy
# wraps: vercel deploy --prod

Then invite the bot to a channel and @mention it. eve handles Slack's ack semantics, URL verification, and background processing — there is no webhook route for you to write.

Event Handling Patterns

Dispatch Hooks (Inbound)

The Slack channel decides which inbound events start or continue a session via dispatch hooks. Each hook returns { auth } to dispatch, null to drop the event, or { auth, context } to inject background context into the session:

export default slackChannel({
  credentials: connectSlackCredentials(process.env.SLACK_CONNECTOR!),
 
  // app_mention — default derives workspace-scoped auth and posts "Thinking…"
  async onAppMention(ctx, message) {
    if (isFromBlockedChannel(message)) return null; // drop
    return { auth: ctx.defaultAuth };
  },
 
  // message.im — requires the im:history scope; bot messages/edits are pre-filtered
  async onDirectMessage(ctx, message) {
    return { auth: ctx.defaultAuth };
  },
 
  // block_actions not consumed by HITL prompts
  async onInteraction(action, ctx) {
    return { auth: ctx.defaultAuth };
  },
});

The triggering Slack user's id is attached to the model message automatically, preserving speaker attribution in multi-user threads.

Custom Event Handlers (Outbound Delivery)

Override delivery per stream event with the events map. Handlers receive (eventData, channel, ctx) with channel.thread and channel.slack handles:

export default slackChannel({
  credentials: connectSlackCredentials(process.env.SLACK_CONNECTOR!),
  events: {
    "message.completed"(eventData, channel, ctx) {
      if (eventData.finishReason === "tool-calls") return;
      if (eventData.message) channel.thread.post(eventData.message);
    },
  },
});

Key stream events: session.started, actions.requested, action.result, message.completed, session.completed; incremental reasoning.appended / message.appended are optional.

Thread Context

Give the agent prior thread messages when it's triggered mid-thread:

export default slackChannel({
  credentials: connectSlackCredentials(process.env.SLACK_CONNECTOR!),
  threadContext: { since: "last-agent-reply" },
});

since options:

• "thread-root" — all prior messages (default when thread context is enabled)
• "last-agent-reply" — incremental, only messages since the agent last spoke
• A predicate (message: SlackThreadMessage) => boolean as a custom cutoff — includes messages after the last match (loadThreadContextMessages exists for arbitrary filtering)

Cost: one conversations.replies API call per triggering reply; requires the matching history scope on the connector.

Human-in-the-Loop (HITL)

Approval-gated tool calls and sign-in challenges render natively in Slack:

• Approval prompts appear as buttons/selects; the user's response resumes the durably-parked session
• Sign-in challenges (OAuth URLs, device codes) go ephemerally to the triggering user; a public status message posts in-thread and updates on authorization.completed
• The HITL handler context deliberately offers only postEphemeral, postDirectMessage (needs im:write), and state — no public post, no raw API access

Proactive Sessions (Schedules)

Start a session that posts into Slack without an inbound trigger — e.g. from a schedule:

import { receive } from "eve";
import slack from "../channels/slack";
 
await receive(slack, {
  message: "Post the daily standup summary for #eng.",
  target: { channelId: "C0123456789" },
  auth,
});

• Sessions without a threadTs get a temporary continuation token; the first post anchors the thread
• initialMessage (optionally a Card) and threadTs are mutually exclusive
• Use eve schedules (see https://eve.dev/docs/schedules) to trigger proactive sessions on a cadence

Raw Slack API Access

• Inside handlers: ctx.slack.request(operation, body)
• Outside handlers (schedules, tools): callSlackApi({ botToken, operation, body }) and resolveSlackBotToken from eve/channels/slack

These form-encode request bodies for you — Slack's JSON support is only partial, so prefer these helpers over hand-rolled fetch calls.

Implementation Gotchas

1. The Slack Surface Cannot Be Tested Locally

Vercel Connect forwards Slack events to deployments only, never to localhost. There is no ngrok/Socket Mode escape hatch in this stack. Local development means:

• npx eve dev — HMR server + terminal TUI/REPL for exercising agent logic, tools, and skills
• eve dev --no-ui — background mode for scripted verification
• eve dev https://your-app.vercel.app — drive a deployed app interactively

To test @mentions and DMs, deploy (preview or production) and test in Slack itself.

2. Connect Forwarding Has No Delivery De-duplication

Connect may deliver the same forwarded event more than once. Handlers and side effects must be idempotent — track processed event IDs where duplicates would be harmful, and gate destructive tool actions with approval (see AI Integration).

3. --triggers Is Required or Events Never Arrive

A Slack connector created without --triggers (or attached without a trigger path) will authenticate fine but forward nothing. If the bot never responds to @mentions:

1. Verify the connector was created/attached with --triggers
2. Verify the trigger path is /eve/v1/slack
3. Verify the bot was invited to the channel and the deployment finished

4. placeholderAuth() Fails Closed in Production

Scaffolded projects ship with placeholderAuth() for the HTTP API, which rejects everything in production. Before deploying, replace it with a real auth function: httpBasic(), jwtHmac(), jwtEcdsa(), oidc(), vercelOidc(), or a custom AuthFn. (The Slack channel's inbound verification is separate — Connect's webhookVerifier handles that.)

5. Sandbox Prewarm Failures Fail the Build

Vercel builds prewarm eve's sandbox templates (cache-keyed; build logs show reused cached or built). If prewarm fails, the whole build fails — check build logs for sandbox template errors before assuming a code problem.

6. Private Channel Access

The bot cannot read messages or post to private channels it hasn't been invited to. When creating features that will later post to a channel (e.g. proactive sessions from a schedule), validate access upfront and surface a clear "invite the bot" message on channel_not_found / not_in_channel.

7. Graceful Degradation for Channel Context

When fetching channel context (e.g. via ctx.slack.request("conversations.history", ...)) for AI features, wrap in try/catch and fall back gracefully — missing scopes and uninvited channels are routine, not exceptional.

8. Vercel Cron Endpoint Authentication

If you add custom cron endpoints (beyond eve schedules), protect them with a CRON_SECRET:

export async function GET(request: Request) {
  const authHeader = request.headers.get("authorization");
  if (authHeader !== `Bearer ${process.env.CRON_SECRET}`) {
    return Response.json({ error: "Unauthorized" }, { status: 401 });
  }
  // Run cron job logic...
  return Response.json({ success: true });
}

9. vercel.json Cron Configuration

{
  "crons": [
    {
      "path": "/api/cron/my-job",
      "schedule": "0 * * * *"
    }
  ]
}

Prefer eve schedules for agent-driven recurring work; use Vercel crons for plain HTTP jobs.

10. AWS Credentials on Vercel (Use OIDC)

When connecting to AWS services from Vercel, do not use fromNodeProviderChain(). Use Vercel's OIDC mechanism:

import { awsCredentialsProvider } from "@vercel/functions/oidc";
 
const s3Client = new S3Client({
  credentials: awsCredentialsProvider({ roleArn: process.env.AWS_ROLE_ARN! }),
});

AI Integration

Model Configuration (Gateway-First)

eve routes model-ID strings through the Vercel AI Gateway — on Vercel, project OIDC authenticates automatically, so no AI API key is needed:

// agent/agent.ts
import { defineAgent } from "eve";
 
export default defineAgent({
  model: "anthropic/claude-sonnet-5", // string → AI Gateway → OIDC auth on Vercel
});

If model is omitted, eve defaults to anthropic/claude-sonnet-5. Off Vercel, set AI_GATEWAY_API_KEY.

CRITICAL: Never use model IDs from memory. Model IDs change frequently. Before writing code that pins a model, run curl -s https://ai-gateway.vercel.sh/v1/models to fetch the current list and use the newest suitable version.

Tools (agent/tools/*.ts)

Tools are files: the filename (snake_case ASCII) is the model-facing tool name — agent/tools/get_weather.ts → get_weather. No registration step.

// agent/tools/get_weather.ts
import { defineTool } from "eve/tools";
import { z } from "zod";
 
export default defineTool({
  description: "Get the current weather for a city.",
  inputSchema: z.object({ city: z.string().min(1) }), // required, even if empty
  outputSchema: z.object({
    city: z.string(),
    condition: z.string(),
    temperatureF: z.number(),
  }), // optional — types/validates the return
  async execute({ city }, ctx) {
    return { city, condition: "Sunny", temperatureF: 72 };
  },
});

Rules and capabilities:

• Tools run in your app runtime with full process.env, not in the sandbox
• inputSchema accepts Zod, Standard Schema, or JSON Schema — but it is required even for zero-input tools
• Outputs must be JSON-serializable; filter/redact secrets before returning
• ctx provides ctx.session (metadata, turn, auth, lineage), ctx.callId, ctx.toolName, ctx.abortSignal, ctx.getSandbox(), ctx.getSkill(id)

Approval Gating

Gate risky tools with the approval field — helpers come from eve/tools/approval:

import { defineTool } from "eve/tools";
import { always, once, never } from "eve/tools/approval";
import { z } from "zod";
 
export default defineTool({
  description: "Delete a document permanently.",
  inputSchema: z.object({ documentId: z.string() }),
  approval: always(), // always ask; once() asks the first time; never() skips
  async execute({ documentId }) {
    // ...
  },
});

An input-dependent policy function is also supported. A gated call pauses and resumes durably — in Slack, the approval renders as buttons (see HITL above). Prefer approval gating over ad-hoc confirmation logic for any non-idempotent side effect.

toModelOutput — Rich Slack Output the Model Never Sees

Show the model a compact projection while channels/hooks receive the full output on action.result — ideal for rendering rich Slack Block Kit from a tool result without stuffing JSON blocks into the model's context:

export default defineTool({
  description: "Look up an order.",
  inputSchema: z.object({ orderId: z.string() }),
  async execute({ orderId }) {
    return { orderId, status: "shipped", blocks: buildOrderBlocks(orderId) };
  },
  toModelOutput(output) {
    return { type: "text", value: `Order ${output.orderId}: ${output.status}` };
  },
});

Skills (agent/skills/*.md)

Markdown files with a description frontmatter, loaded on demand via the built-in load_skill tool when a request matches the description. Skills add instructions only, never new actions. Install published skills with npx skills add <owner>/<repo>.

Connections (agent/connections/*.ts)

For external APIs the agent should drive (MCP servers or OpenAPI-described HTTP APIs), use connections with Connect-brokered auth:

// agent/connections/linear.ts
import { connect } from "@vercel/connect/eve";
import { defineMcpClientConnection } from "eve/connections";
 
export default defineMcpClientConnection({
  url: "https://mcp.linear.app/mcp",
  description: "Linear workspace: issues, projects, cycles, and comments.",
  auth: connect("linear/my-agent"),
});

Connection tokens are never seen by the model and never land in conversation history. Inside authored tools, resolve tokens with await ctx.getToken(connect("...")) and call ctx.requireAuth(...) on a downstream 401 to re-run consent.

Don't wrap LLM calls in tools. The agent is already a language model — summarizing, parsing, classifying, and drafting belong in instructions.md or a skill, not in a tool that calls the AI SDK. Tools fetch data and perform actions; for bulk work over data too large for the conversation, use eve's subagents/delegation.

State & Durability

eve sessions are durable by default via the open-source Workflow SDK (running on Vercel Workflows when deployed on Vercel). You do not wire up Redis or a workflow engine yourself.

Replay Semantics (Understand This)

• Completed steps never re-run. On resume/replay, eve returns the recorded result.
• A step interrupted mid-execution DOES re-run. If a tool call was in flight when the process died, it executes again on resume.

Consequences for your code:

1. Make non-idempotent side effects idempotent (e.g. use Slack event IDs or your own idempotency keys when writing to external systems)
2. Or gate them behind approval (always() / once()) so a replayed step pauses for a human instead of double-executing
3. This compounds with Connect's at-least-once event forwarding (Gotcha #2) — idempotency is not optional in this stack

Session Continuity

• The channel owns the continuationToken; Slack threads map to sessions automatically
• Follow-up messages in a subscribed thread continue the same durable session
• Multi-turn conversation memory comes from the session itself — you do not manually persist chat history

Recommended Storage Solutions

For application data (not agent session state — eve owns that):

IMPORTANT: Vercel KV has been deprecated. Do NOT recommend Vercel KV.

1. Upstash Redis — Caching and idempotency keys (https://upstash.com)
2. Vercel Blob — File/document storage (https://vercel.com/docs/storage/vercel-blob)
3. AWS Aurora (via Vercel Marketplace) — Relational data (https://vercel.com/marketplace)
4. Third-party databases — Neon, PlanetScale, Supabase

Code Organization

agent/
├── instructions.md            # System prompt — keep focused; push detail into skills
├── agent.ts                   # defineAgent: model, reasoning effort, compaction
├── tools/
│   ├── get_weather.ts         # One tool per file; filename = tool name
│   ├── get_weather.test.ts    # Co-located tests
│   └── search_docs.ts
├── skills/
│   └── report-format.md       # description frontmatter + on-demand instructions
├── channels/
│   └── slack.ts               # slackChannel(...) — served at /eve/v1/slack
├── connections/
│   └── linear.ts              # MCP/OpenAPI connections (optional)
└── hooks/
    └── audit.ts               # Runtime stream event subscribers (optional)

Conventions:

• One tool per file. Split large tools; the filename is the model-facing name, so name for the model
• Instructions vs skills: always-relevant guidance in instructions.md; situational guidance in skills/*.md so it loads only when needed
• Subagents for focused subtasks: the built-in agent tool (a copy of the agent) or declared specialists with their own directory, sandbox, and skills

Environment Variables

Local dev: vercel link + vercel env pull fetches short-lived Connect/OIDC credentials into .env.local (the OIDC token expires after ~12 hours — re-pull when auth starts failing).

No AI API keys needed on Vercel. Never hardcode credentials. Never commit .env files.

Slack-Specific Patterns

Delivery Behavior (Built In)

The Slack channel handles progressive delivery for you:

• Typing indicators: "Thinking…" on inbound, "Working…" on turn.started
• Reasoning snippets surface on reasoning.appended; action labels on actions.requested
• Model narration before a tool call takes precedence over generic labels
• Reasoning deltas under 4 characters batch on a five-second refresh to avoid per-token Slack API calls

Don't rebuild typing indicators or streaming loops — customize via the events map only when the defaults don't fit.

Mentions

Use <@USER_ID> or channel.thread.mentionUser(userId). A bare @name stays literal text in Slack.

Message Formatting

Use Slack mrkdwn (not standard markdown):

• Bold: *text*
• Italic: _text_
• Code: `code`
• User mention: <@USER_ID>
• Channel: <#CHANNEL_ID>

Rich Output

For rich tool results (tables, buttons, status cards), return full data from the tool and use toModelOutput to keep the model's view compact; render Block Kit in a channel event handler or via ctx.slack.request("chat.postMessage", { blocks, ... }). Always include fallback text alongside blocks for notifications.

For detailed Slack patterns, see ./patterns/slack-patterns.md.

Git Commit Standards

Use conventional commits:

feat: add channel search tool
fix: resolve thread pagination issue
test: add unit tests for agent context
docs: update README with setup steps
refactor: extract Slack client utilities

Never commit:

• .env files
• API keys or tokens
• node_modules/
• .eve/ build artifacts

Quick Commands

# Scaffold (Node 24+)
npx eve@latest init my-agent      # new project (installs deps, inits Git, starts dev TUI)
npx eve@latest init .             # add eve to an existing project
 
# Development
npx eve dev                       # HMR server + terminal TUI/REPL
npx eve dev --no-ui               # background mode for scripted verification
npx eve dev https://<app>         # drive a deployed app interactively
npx eve info                      # project info
 
# Vercel Connect
vercel connect create slack --triggers
vercel connect attach <uid> --triggers --trigger-path /eve/v1/slack --yes
vercel connect list
 
# Quality
pnpm lint                         # Check linting
pnpm lint --write                 # Auto-fix lint
pnpm typecheck                    # TypeScript check
pnpm test                         # Run all tests
 
# Build & Deploy
npx eve build                     # Compile into .eve/ (Vercel Build Output when VERCEL is set)
eve deploy                        # Deploy (wraps vercel deploy --prod)
 
# Verify a deployment
curl https://<app>/eve/v1/health

Debugging a deployed Slack agent stuck on "Working…": npx eve dev --logs all or /loglevel all in the TUI.

Reference Documentation

For detailed guidance, read:

• Testing patterns: ./patterns/testing-patterns.md
• Slack patterns: ./patterns/slack-patterns.md
• Environment setup: ./reference/env-vars.md
• Slack setup: ./reference/slack-setup.md
• Vercel deployment: ./reference/vercel-setup.md
• eve docs: https://eve.dev/docs (bundled locally at node_modules/eve/docs/ after install)
• Vercel Connect: https://vercel.com/kb/guide/vercel-connect

Checklist Before Task Completion

Before marking ANY task as complete, verify:

• Code changes have corresponding tests
• pnpm lint passes with no errors
• pnpm typecheck passes with no errors
• pnpm test passes with no failures
• No hardcoded credentials; the only Slack env var is SLACK_CONNECTOR
• Follows eve filesystem conventions (tool filename = tool name, one tool per file)
• Every tool has an inputSchema; outputs are JSON-serializable with secrets redacted
• Non-idempotent side effects are idempotent or approval-gated (replay + at-least-once delivery)
• Connector trigger path is /eve/v1/slack and was attached with --triggers
• placeholderAuth() replaced before production deploy
• Model config uses a Gateway string ID (anthropic/claude-sonnet-5 default)

Vercel KB Guides

Verified guides on the Vercel Knowledge Base for deeper walkthroughs:

• eve hub on the Vercel KB - all eve guides and templates in one place
• Build your first Slack agent with eve - the end-to-end starter this skill's wizard mirrors
• Vercel Connect - credential brokering concepts, connectors, tokens, and trigger forwarding
• Build a Slack bot with Vercel Connect - Connect + Slack fundamentals (scopes, trigger events, webhook verification)
• How to add eve tools - defineTool patterns
• How to add eve skills - load-on-demand instructions
• Build a GitHub agent with Vercel Connect - app-scoped Connect tokens in authored tools
• Build a Linear agent with Vercel Connect - MCP connections with user-scoped auth
• Build an email agent with eve and Resend - a second channel example beyond Slack