rome-os/rome/composio-cli
Claude Code and agent skill by rome-os/rome
- community
Checking that the SKILL.md file is still in the repository.
Help users operate the published Composio CLI to find the right tool, connect accounts, inspect schemas, execute tools, subscribe to trigger events with `composio listen`, script workflows with `composio run`, and call authenticated app APIs with `composio proxy`. Use when the user asks how to do something with `composio`, wants to run a known tool slug, needs to discover a slug with `composio search`, fix a missing connection with `composio link`, inspect tool inputs with `--get-schema` or `--dry-run`, troubleshoot top-level CLI flows, or explicitly needs `composio dev` guidance.
Install
npx vskill@latest install rome-os/rome/composio-cliSKILL.md
Composio CLI
Preflight: Is The User Logged In?
Before running any toolkit command, confirm the user is logged in. If they are not, stop and tell them to log in — do not improvise.
If whoami fails (no Composio session), the user is logged out. Stop here. Do not:
- run
composio loginyourself, - look for an alternate toolkit, slug, or
proxyroute, - ask the user to paste a key, or
- attempt any other workaround.
Respond directly to the user with one message: they need to log in to Composio first, via the "Log in with Composio" action on the Composio app's dashboard. Then end your turn and wait — the request cannot proceed until they have logged in.
Only once whoami succeeds do you continue to the workflow below.
Connection Invariant: One Account Per Toolkit
Rome operates exactly one connected account per toolkit. This is load-bearing because execute, proxy, and run have no flag to choose an account — they act on the toolkit's single connected account, and when more than one exists Composio silently resolves a non-deterministic default. You cannot tell, and cannot control, which account a tool ran as. A second Gmail means a "send email" might go out from the wrong account with no signal.
To keep every tool call deterministic:
- Never create a second account for a toolkit. Do not pass
--aliasto
composio link to add an additional account, and do not enable the multi_account experimental feature. The alias buys you nothing on execution — there is no --alias/--account on execute — it only lets you accumulate ambiguous accounts.
- Check before linking.
composio link <toolkit> --listshows existing
accounts. If one is ACTIVE, you are connected — retry execute, do not link again.
- If `--list` shows more than one account for a toolkit, stop and surface it.
Tell the user the toolkit has duplicate accounts and tool calls are ambiguous until the extras are removed. Do not guess which one to use.
Default Workflow
- Start with
composio execute <slug>whenever the slug is known. - If several independent tool calls must happen at once, use
composio execute -p/--parallelwith repeated<slug> -d <json>groups. - If
executesays the toolkit is not connected, runcomposio link <toolkit>and retry. - If the arguments are unclear, run
composio execute <slug> --get-schemaor--dry-runbefore guessing. - Reach for
composio search "<task>"only when the slug is unknown.searchaccepts one or more queries, so batch related discovery work into a single command when useful.
execute - Run A Tool
Use execute when the tool slug is already known.
Inspect required inputs without executing:
Preview safely:
Pass data from a file or stdin:
Upload a local file:
Run independent tool calls in parallel:
Key flags:
--get-schema: Inspect required arguments without executing the tool.--dry-run: Preview the request shape without performing the action.--file: Inject a local file path into a tool that exposes exactly one uploadable file argument.--parallel: Execute multiple independent tool calls in the same invocation.
There is no flag to pick an account — execute uses the toolkit's single connected account (see "Connection Invariant" above). Keep one account per toolkit so this is unambiguous.
--fileonly works when the tool exposes a single uploadable file input. Otherwise use explicit-dJSON.
search - Find The Slug
Use search only when the tool slug is not already known.
- Batch related discovery work into one
searchinvocation, then move back toexecuteonce the correct slugs are known.
link - Connect An Account
Use link when execute reports that a toolkit is not connected, or when the user explicitly wants to authorize an account.
Key flags:
--list: List the toolkit's existing connected accounts instead of linking. Run
this first — if an account is already ACTIVE, retry execute rather than linking.
--alias: Names an additional account for a toolkit (requires the
multi_account experimental feature). Avoid it — Rome keeps one account per toolkit, and extra accounts make execute ambiguous (see "Connection Invariant").
- Retry the original
executecommand after linking succeeds.
proxy - Raw API Access
Use proxy when a toolkit supports a raw API operation that is easier than finding a dedicated tool slug.
run - Scripting, LLMs, and Programmatic Workflows
For programmatic calls, loops, output plumbing, or anything beyond a single tool call, prefer composio run.
composio run executes an inline ESM JavaScript/TypeScript snippet with authenticated execute(), search(), proxy(), and the experimental experimental_subAgent() helper pre-injected. No SDK setup required.
Chain multiple tools:
Fan out with Promise.all:
Feed tool output into an LLM and get structured JSON back:
- Use top-level
execute --parallelinstead when the user only needs a few independent tool calls and does not need script logic.
Bundled Scripts
Reusable scripts ship in rome_apps/composio/scripts/ and are invoked via composio run -f.
gmail-parse-markdown.ts — Gmail → Markdown
GMAIL_FETCH_EMAILS returns each body as raw HTML — a single 50-message page is typically ~200k tokens, dominated by HTML markup and tracking URLs. This script fetches messages and converts each body to clean Markdown using @mozilla/readability (with a turndown fallback when readability mis-fires on transactional emails). Cuts a 50-message page to ~3-15k tokens depending on content.
Reach for it any time the user wants Gmail content fed to an LLM, summarized, or otherwise processed as text — not when they need raw HTML.
Flags:
--query "<gmail-query>"— Gmail search query (e.g."is:unread newer_than:1d")--max <N>— max_results (default 10, max 500)--page-token <token>— pagination cursor from a priornextPageToken--include-links— keep inline links (default strips URLs but preserves anchor text; ~3x token savings)--raw— also return the originalmessageTextalongsidemarkdown(debugging only)
Output is JSON on stdout: { messages: [{ messageId, threadId, sender, subject, markdown, bodyFormat, ... }], nextPageToken, resultSizeEstimate, options }.
Auth
When whoami fails the user is logged out. Do not run `composio login` on their behalf and do not work around the missing session. Tell the user to log in via the dashboard and stop — see "Preflight: Is The User Logged In?" above.
The Composio Rome app reads the key live from the CLI session (~/.composio/user_data.json) — that session is the single source of truth; the app stores no copy of its own. Its dashboard exposes Composio login as one explicit browser action: "Log in with Composio" (or "Re-authorize" once logged in) opens Composio in a new tab to authorize Rome, which writes the issued key into the CLI session. The dashboard deliberately hides the CLI mechanics — the guardian just sees the button.
Because the app reads the CLI session directly, a composio login you (the agent) run in this terminal is picked up automatically — there is no separate "import" step.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingrome_apps/connector/src/skills/composio-cli/SKILL.md
Popularity
- Repository stars
- 672
- Repository forks
- 57
- 7-day trend score
- 33
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/rome-os/rome/composio-cli)