Agent skill · Audited from source

Slack

skill · vellum-ai/vellum-assistantMIT licenceAudited 14 Oct 2026 · source 19bc86343b63

★ 1,425pushed 1 day ago
✓SAFEGrade B · 89/100 · 14 Oct 2026 · 19bc863
1,425repo stars · whole vellum-ai/vellum-assistant

Is it safe? Yes — SAFE, grade B (89/100). We read this skill's source at commit 19bc86343b63 and found no filesystem, network or shell access and no hard-coded secrets. What it does: Send, read, and manage Slack messages

Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
More
Cite

Cite this page

ModelsAtlas. “Slack (vellum-ai/vellum-assistant)”. Retrieved 14 Oct 2026. https://modelsatlas.com/skills/vellum-ai/slack

2 skills are called “Slack”. You are viewing vellum-ai/vellum-assistant.

Trust audit

Audit 0.4.1 · source read 14 Oct 2026

Trust score89/100✓SAFE B
0 critical0 high0 medium5 low0 info

SAFE means nothing in the source contradicts what the skill says it does. Grade B is shared by 13,985 of 14,816 audited skills.

What the source does

  • FilesystemNone observedNo reads or writes in the source
  • NetworkNone observedNo outbound calls in the source
  • ShellNone observedNo commands run in the source
  • DependenciesPinnedNothing floats
  • SecretsNone foundNo keys in source

Five audit layers

LayerWhat we checkResultFindings
L0 ProvenanceRepo owner, licence, archive status, activityPASS5
L1 Static sourceSecrets, path traversal, supply chain, hidden codePASS0
L2 InstructionsPrompt injection, authority framing, hidden directivesPASS0
L3 InventoryThe files and declarations match the sourcePASS0
L4 SandboxRun in isolation and watch behaviourSKIPPED why0

Gates applied: no behavioural pass.

Findings (5)

Audit 0.4.1 · source sha 19bc86343b63 · 14 Oct 2026Report an issue or request a re-scan

Provenance

Who published it, and what the verdict is pinned to

  • Publishervellum-ai GitHub organizationOwner of the repository this skill is read from
  • Source foldervellum-ai/vellum-assistant/skills/slackThe exact folder we audited
  • LicenceMITAs the repository states it
  • Commit pinned for this verdict19bc86343b63A new audit runs when the source changes
  • Repository statusNot archivedLast push 1 day ago
  • Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

What it tells the agent

Full source on GitHub ↗

Its description: Send, read, and manage Slack messages

---
name: slack
description: Send, read, and manage Slack messages
compatibility: "Designed for Vellum personal assistants"
metadata:
  icon: assets/icon.svg
  emoji: "💬"
  vellum:
    category: "messaging"
    display-name: "Slack"
---

You help users interact with their Slack workspace.

**Sending text is its own command: `assistant channels send`.** It posts through the same path a reply takes, so the message is threaded where you name a thread, rendered the way your other messages are, and recorded in the chat's conversation once Slack acknowledges it. That record is what lets you see later what you said, and what lets a reaction or an edit on that post find it again.

**Everything else is the Slack Web API, through one of two doors.** As the assistant's own bot: `assistant channels request slack <path>`, which resolves the bot credential from the channel so you never name a provider (`assistant oauth request --provider slack_channel` is the same bot door under its provider name). As the person who connected the `slack` integration: `assistant oauth request --provider slack <path>`, for `search.messages` and anything that needs the installer's reach. Reading, reactions, opening a DM, uploading a file, and any method with no first-class command go through them. Use relative Slack API method paths such as `/conversations.history`; the provider supplies the Slack host.

A message posted with `/chat.postMessage` reaches Slack but is not recorded in the chat's conversation, so `recall` will not find it (reading Slack back will) and nothing can resolve a reaction to it. Use the send command for text, and reach for `chat.postMessage` only for a shape the send command does not carry, such as Block Kit.

**In a Slack turn, the reply is the door.** When you are already answering in a chat, just reply; that reply is delivered and recorded for you. The send command is for posting somewhere you are not, or at a time nobody asked.

## Which provider to pass

Two Slack credentials can exist. They act as different identities and reach different things, so the right one depends on the operation, not only on which is configured.

| Provider        | Sends                      | Acts as       | Reaches                                           |
| --------------- | -------------------------- | ------------- | ------------------------------------------------- |
| `slack_channel` | the bot token              | the assistant | channels the bot has joined                       |
| `slack`         | the installer's user token | that person   | channels that person is in, and `search.messages` |

**Posting: `slack_channel`.** The send command always posts as the bot, which is the assistant's own identity, and takes no provider. The distinction still matters for a Web API post: a message sent through `slack` arrives from the person who connected it, not from the assistant. Where that is the only credential present, say so before posting rather than posting as them silently.

**`search.messages`: `slack` only.** It is a user-token method. `--provider slack_channel` sends the bot token even on an install that stored a user token, so it cannot serve search at all.

**Everything else: whichever is present**, preferring `slack_channel` when both are. Reach differs rather than one being better: the bot sees channels it was invited to, the user token sees channels that person belongs to.

A workspace has only `slack` when Slack was connected as an integration without running the setup wizard, which means no bot. `oauth status <provider>` reports whether a given one holds a connection. Passing a provider that holds none fails with a not-configured error rather than falling back on its own.

## Resolution Scripts

Use these scripts to resolve Slack channel and user names to IDs. Results are cached locally so repeated lookups are free (no API calls).

| Command                                                          | Description                                      |
| ---------------------------------------------------------------- | ------------------------------------------------ |
| `bun skills/slack/scripts/slack-resolve.ts channel <name>`       | Resolve a channel name to its ID                 |
| `bun skills/slack/scripts/slack-resolve.ts user <name-or-email>` | Resolve a user display name or email to their ID |
| `bun skills/slack/scripts/slack-resolve.ts channels [--refresh]` | List all cached channels, or refresh the cache   |

All scripts return JSON:

- **Success**: `{ "ok": true, "data": { "id": "C...", "name": "general", ... } }`
- **Failure**: `{ "ok": false, "error": "..." }`

The cache is stored locally under `$VELLUM_WORKSPACE_DIR/data/slack-skill/`. On first use the script fetches all channels/users from Slack and caches them. Subsequent lookups read from the cache with no API calls. Pass `--refresh` to force a refresh.

## Sending a message

The send command names the channel, the chat, and the text. It takes no provider and no token.

```bash
assistant channels send slack C0123456789 --text "Hello from the assistant!"
```

Add `--thread <ts>` to post inside a thread, `--plain` to send the text verbatim instead of the channel's rich rendering, and `--json` for a machine-readable result naming every message id Slack acknowledged, plus the conversation the post was recorded in when a record was written.

The command refuses a channel it cannot address before anything is sent, and reports a failure rather than a success when Slack does not acknowledge the post. If it reports that the outcome is unknown, the message may still have gone out: check the chat before sending again.

## Making Slack API Calls

Two commands reach the Slack Web API, and both send the token from inside the assistant. `assistant channels request slack <path>` acts as the bot, resolving the bot credential from the channel id. `assistant oauth request --provider <key> <path>` names the credential by provider: `slack_channel` is 

Excerpt, first 6,001 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

Install

Pinned to the audited commit 19bc863

Download folder ↗
git clone https://github.com/vellum-ai/vellum-assistant
git -C vellum-assistant checkout 19bc86343b63

The author documents no install command, so this uses the repository directly. Our verdict covers commit 19bc86343b63 only.

Works in

The author has not said which agents this skill supports, and we have not tested it in any.

Audit history

Every audit, every source change

  1. ✓SAFE B 89 / 100

    first audit · source 19bc86343b63

README badge

For maintainers · shows the latest audit

ModelsAtlas: SAFE BMODELSATLASSAFE · B
[![ModelsAtlas: SAFE B](https://modelsatlas.com/skills/vellum-ai/slack/badge.svg)](https://modelsatlas.com/skills/vellum-ai/slack/)

The badge shows our verdict and grade for this skill, and links back to this page. It changes when the audit does.

Add this badge to your README

[![ModelsAtlas: SAFE B](https://modelsatlas.com/skills/vellum-ai/slack/badge.svg)](https://modelsatlas.com/skills/vellum-ai/slack/)

Questions about Slack

What does the Slack skill do?
An AI Assistant that’s easy to setup, does your work 24/7, knows your preferences and gets better over time.
Is Slack safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can Slack access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
How current is this page?
The grade is for one exact copy of the source (19bc86343b63), read on 2026-10-14. The repository is watched, and a new audit runs when it changes — this is the first audit.

Sources

Spotted something wrong? Report a correction · Methodology

Keep going in your AI

Opens with this page as the source. We add nothing else to the prompt.