Use when a developer works with the Slack CLI (slack command) to create, run, or manage a Slack app from the terminal: logging in or authenticating (slack login), adding a team or switching workspaces, running locally (slack run), deploying, editing the app manifest, calling Web API methods (slack api), or searching Slack docs (slack docs search).
Use when a developer works with the Slack CLI (slack command) to create, run, or manage a Slack app from the terminal: logging in or authenticating (slack login), adding a team or switching workspaces, running locally (slack run), deploying, editing the app manifest, calling Web API methods (slack api), or searching Slack docs (slack docs search).
Slack CLI
Use the Slack CLI to create, run, and manage Slack apps — including calling Web API methods directly and searching Slack developer documentation from the terminal.
For initial setup (sandbox creation, project scaffolding from templates), use the slack:create-slack-app skill instead.
Step 1: Detect the Slack CLI
Resolve the path to the public Slack CLI before running any other command. We refer to the resolved value as SLACK_CMD — substitute it everywhere below. Try 1a → 1b → 1c in order; do not skip ahead.
1a. Standard install path (preferred — no verification needed)
The official install scripts place the binary at a fixed per-user path. If the file exists, use it directly: only the install script writes to that path.
macOS / Linux: ~/.slack/bin/slack
Windows: %USERPROFILE%\AppData\Local\slack-cli\bin\slack.exe (or %USERPROFILE%\.slack-cli\bin\slack.exe as a fallback)
Then re-run 1a — the install script will have written ~/.slack/bin/slack.
Common mistakes: Don't use which slack to discover the binary — which resolves any shell alias and defeats the point of 1a. In Git Bash on Windows, use the POSIX probe form, not PowerShell.
Step 2: Command Discovery via Help
Always run SLACK_CMD <command> --help before constructing a command you have not used in the current session. Do not guess at flags — the help output is the source of truth.
SLACK_CMD help — lists all available command groups
SLACK_CMD <command> --help — shows subcommands, flags, and usage examples
Resolving --app and --team values
When a command requires --app or --team:
App ID: Run SLACK_CMD app list from the project directory to see installed apps and their IDs.
Team ID: Run SLACK_CMD auth list to see authenticated workspaces and their team IDs.
Use --output=text for concise terminal-readable results. Use this when you are already running CLI commands and want to:
Check a Slack feature before implementing it
Verify API behavior or required scopes
Look up Block Kit elements, event types, or method parameters
Step 4: Calling Web API Methods (slack api)
Call any Slack Web API method directly. Run SLACK_CMD api --help for full details and examples.
Parameters are passed as positional key=value pairs (NOT --key=value flags):
SLACK_CMD api chat.postMessage channel=C0123456789 text="Hello from the CLI"
Important distinction: --team, --token, --json, and --data are meta-flags (prefixed with --). API method parameters use positional key=value syntax without dashes.
Slack auth is per-team, not a single boolean. Run the seamless login flow below whenever any of these is true:
The developer is not authenticated to any team yet.
The developer wants to add a new team / workspace / sandbox, even if SLACK_CMD auth list already shows other teams.
The developer asks to log in, switch teams, or re-authenticate.
SLACK_CMD auth list showing other teams is not a reason to skip login — those are different teams. Ask the developer which team they want, then run the flow.
Inspect existing auth (optional)
SLACK_CMD auth list
Use this to show the developer which teams are already authenticated, or to confirm a successful login. Do not treat a non-empty list as "auth complete" when the developer asked to log in to a new team.
Run the seamless login flow
The agent drives this end-to-end — no separate terminal window, no browser confirmation.
SLACK_CMD login --no-prompt makes the CLI emit a single-use ticket and exit immediately instead of waiting on stdin. Slack itself renders the challenge code inside a workspace modal when the developer sends the /slackauthticket slash command — there is no browser step. The agent submits the challenge back to the CLI to complete login.
1. Start login and capture the ticket
SLACK_CMD login --no-prompt
The CLI prints a /slackauthticket <ticket> slash command and exits. Capture the ticket — you will need it in step 4. Sample output:
📋 Run the following slash command from any Slack channel in the workspace
you'd like to authenticate
/slackauthticket eyJ0eXAiOiJKV1QiLCJh…
? Slack will then show you a challenge code. Submit it via:
slack login --ticket <ticket> --challenge <code>
2. Hand the slash command to the developer
Show the full /slackauthticket … line and ask the developer to paste it into the message box of the Slack workspace they want to authenticate, then send it. Slack responds with a modal containing a short challenge code (e.g. JDt1IK7X).
3. Collect the challenge code
Use AskUserQuestion to ask the developer for the challenge code shown in the Slack modal. Wait for their answer — do not guess or default.
On success the CLI returns the team name and ID. Verify with SLACK_CMD auth list and report the team back to the developer.
Troubleshooting: tickets are single-use and time-limited. If step 4 fails with an invalid/expired ticket or wrong challenge, restart from step 1 with a fresh SLACK_CMD login --no-prompt — do not retry the same ticket.
Use --team <team_id> on individual commands to target a specific workspace without switching globally.
Red flags — STOP
If you catch yourself thinking any of these, you are about to regress to the old broken flow:
Rationalization
Reality
"auth list already shows teams, so login isn't needed."
Auth is per-team. The developer asked for a new team — drive the flow.
"slack login needs browser confirmation, so I can't drive it."
False with --no-prompt. The challenge code appears in Slack's modal, not a browser. The agent runs both slack login invocations itself.
"I should tell the developer to run slack login in a separate terminal."
Never. Step 5 is the agent's job from start to finish.
All of these mean: run SLACK_CMD login --no-prompt yourself and follow the four numbered steps above.
Step 6: Running an App Locally (slack run)
Run SLACK_CMD run --help for all available flags.
Resolve the app or team target
Run SLACK_CMD app list from the project directory to check for installed apps:
If an app exists: Use --app=<app_id> in the run command.
If no apps are listed: Use --team=<team_id> instead (get the team ID from SLACK_CMD auth list).
Start the dev server in the background
Run this command as a background process so the developer can continue working:
If an app ID was found:
cd <project-dir> && SLACK_CMD run --org-workspace-grant=all --app=<app_id>
If no apps — use the team ID:
cd <project-dir> && SLACK_CMD run --org-workspace-grant=all --team=<team_id>
Tell the developer the app is running in the background. They can ask:
"What's the status of the dev server?" — to check on it
"Show me the output from slack run" — to see activity logs
"Stop the dev server" — to terminate the process
Step 7: Managing the Manifest
Run SLACK_CMD manifest --help for subcommands (validate, info) and the --source local|remote flag.
Use SLACK_CMD manifest validate before deploying or when something seems wrong with the app configuration.
Step 8: Other Commands
For any other command group (e.g., trigger, datastore, env, collaborator, external-auth, deploy), run SLACK_CMD <command> --help to discover subcommands and flags. Run SLACK_CMD help to see all available command groups.
Notes
SLACK_CMD is a placeholder — always substitute the actual command name resolved in Step 1.
Always run --help before constructing a command you have not used in the current session.
Interactive commands (e.g., slack trigger create without --trigger-def) cannot be run in the background. Tell the developer to run these in a new terminal window. slack login is not in this category — drive it inline using the --no-prompt / --ticket / --challenge flow in Step 5.
slack run runs locally for development. slack deploy deploys to Slack's hosted infrastructure. These are different operations — do not confuse them.