Multica Autopilots
Create, manage, and debug Multica autopilots—automated agent tasks triggered by schedule, webhook, or manual runs.
The evidence shows the skill explicitly warns not to run commands with real side effects (like trigger, delete) for testing, and recommends reading documentation first; it also states webhook tokens must be kept secret and not pasted into comments, logs, docs, or PRs. This indicates awareness of data-flow transparency and sensitive-data handling. However, the skill does not provide user confirmation steps or rollback mechanisms, nor does it clearly scope permissions beyond the allowed-tools. Therefore trust is 12; deductions for lack of explicit confirmation workflow and granular permissions.
The skill provides detailed debug steps and clear error-handling explanations, e.g., for webhook delivery statuses (queued, failed) and the cron preview endpoint. However, it relies on external services (multica CLI and server) without installation or dependency notes, and no test evidence is provided. Under static review, reliability is capped at 10; given missing tests and dependency availability info, score is 8.
The skill clearly defines its application scope (creating, updating, inspecting, triggering, debugging autopilots) and provides trigger conditions (e.g., use trigger only when the user explicitly asks for a manual run). But it lacks non-fit boundaries (e.g., if the multica CLI is not installed or server unreachable) and does not mention Chinese-language support or network reachability. Since it's mostly CLI-based, environment fit is adequate but boundaries are unclear. Score is 10.
The skill has a clear hierarchical structure (Quick start, Core model, CLI commands, Debugging, Side effects) with a referenced supplementary file. However, it lacks versioning, changelog, and maintenance responsibility declarations, and the license is NOASSERTION. The repository has CI and tests, but the skill itself doesn't disclose known limitations. Score is 10.
The skill describes core tasks (create, trigger, debug autopilots) with specific commands and expected outputs. However, it lacks example outputs or verification results, and its marginal value is mainly in preventing misuse but without comparative benchmarks. Under static review, effectiveness is capped at 7; given limited evidence of output completeness, score is 6.
The skill references source files (e.g., references/autopilots-source-map.md), providing auditable paths, and the repository has CI configuration and test files, but the skill itself does not include tests. Under static review, verifiability is capped at 5; with no third-party execution evidence, score is 4.
- Static review cannot verify actual execution; it is recommended to run the multica CLI for read-only operations (e.g., list, get) to confirm command validity.
- The skill does not specify a user confirmation workflow; automatically executing side-effect commands (e.g., trigger, delete) may cause unintended changes; strictly follow user instructions.
- The repository license is NOASSERTION; actual license terms need further verification (e.g., the Multica License's self-hosting restrictions).
- Some core functionality depends on overseas services (e.g., GitHub, Discord); mainland-China users should be aware of network reachability, though the skill is mostly CLI-based, minimizing impact.
What does this skill do, and when should you use it?
This skill provides guidance for creating, updating, inspecting, triggering, and debugging Multica autopilots. Multica is an open-source managed agents platform that turns coding agents into real teammates. An autopilot is a durable automation rule that dispatches work to an agent or a squad's leader when a trigger fires (schedule, webhook, or manual). The skill details the `multica autopilot` CLI commands, execution modes (`create_issue` and `run_only`), trigger setup, webhook handling, and troubleshooting. It also warns about side effects like triggering or deleting, and emphasizes not leaking webhook tokens.
Provides CLI operations for Multica autopilots: list autopilots, get details, create autopilots (with title, description, agent, execution mode), update status (active/paused), add schedule/webhook triggers, view runs, trigger manually, and rotate webhook URLs. It guides users to run read commands first and cautions against testing with destructive commands. It also outlines a step-by-step debugging process for troubleshooting why an autopilot didn't run, including checking squads, agents, runtimes, and webhook delivery status.
- A team wants to schedule recurring tasks like daily standups, weekly reports, or periodic audits
- A developer sets up a webhook-triggered automation (e.g., on GitHub events) to automatically create issues and assign them to an agent
- An operator needs to trigger an autopilot manually and monitor its run status
- A user of a multi-agent platform wants to configure automated workflows for their agents without manual prompting
- A support person troubleshoots why an autopilot didn't run and reviews run failure reasons
What are this skill's strengths and limitations?
- Clear, precise CLI commands for creating and managing autopilots
- Explains the core model (trigger, run, execution modes) for understanding how it works
- Emphasizes side effects and token confidentiality, which are important for safety
- Includes a detailed troubleshooting checklist for non-running autopilots
- One skill among many in a larger repository; the overall README may be more comprehensive than this skill alone
- Requires external Multica CLI and server, implying installation and network access
- Does not include installation steps for the skill itself, only for the CLI
- No test suite or examples provided; evidence is based on documentation only
How do you install this skill?
This skill is a built-in skill in the Multica repository (at server/internal/service/builtin_skills/multica-autopilots/SKILL.md). To use it, place the folder containing this SKILL.md into your skill directory for an Agent Skills-compatible client. You'll need to install the Multica CLI: on macOS/Linux use brew install multica-ai/tap/multica or the install script; on Windows use the PowerShell command. Then run multica setup to configure, authenticate, and start the daemon.
How do you use this skill?
Invoke the skill when needing to create, update, inspect, trigger, or debug Multica autopilots. It guides you through using multica autopilot CLI commands. For example, first run multica autopilot list --output json to see existing autopilots, then create one with multica autopilot create --title "..." --description "..." --agent ... --mode create_issue --output json, and add a schedule trigger with multica autopilot trigger-add <autopilot-id> --kind schedule --cron "0 9 * * *" --timezone Asia/Shanghai --output json. Avoid destructive commands like trigger and delete for testing. Installation steps are not documented in the skill, but the general CLI installation is provided.
How does this skill compare with similar options?
No explicit alternatives; the skill is specific to the Multica platform.