SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsdreamina-cli

dreamina-cli

General & other

View repository

Updated 10-15 21:47

About this skill

Dreamina CLI

⚠️ Maintenance-discontinuation notice: the legacy Dreamina CLI (dreamina, v1.4.18) will, as of November 2026 , cease maintenance. For new tasks, please switch to the Canvas replacement skill dreamina-canvas-cli(installation:npx skills add full-aigc-skills/dreamina-skills --skill dreamina-canvas-cli). The contents of this skill are retained as a historical contract reference.

Use this skill when you need Dreamina (即梦) image or video generation, login, session management, or task history work through dreamina.

即梦 is the Chinese product name of Dreamina. For a generic 即梦 request, select dreamina-canvas-cli; use this frozen skill only for an explicit classic-CLI request.

This skill is intentionally short. Detailed flags and supported values belong to the CLI itself, so always treat dreamina -h and dreamina <subcommand> -h as the primary reference.

Read references progressively:

When to use and boundary

Use this umbrella skill from first-time installation through troubleshooting, or when login, sessions, task history, async follow-up, or command routing spans multiple modalities. Do not use it as a replacement for the four execution skills' mode-specific workflow or for prompt-only authoring.

What this tool is for

dreamina is the local CLI entrypoint for all currently exposed Dreamina (即梦) image and video generation workflows, plus the account/session operations around them.

Use it for:

  • checking or reusing an existing Dreamina login session
  • checking account credit
  • managing sessions with dreamina session
  • clearing the local OAuth login state with dreamina logout
  • submitting image generation tasks
  • submitting video generation tasks
  • querying async task results and downloading result media
  • reviewing saved task history

Default workflow

When using this CLI as an agent:

Step 1: Verify the environment

If dreamina is missing, read the install-to-use reference and obtain explicit authorization before installing or updating software. Otherwise start with dreamina version and dreamina -h, then run dreamina <subcommand> -h before real use.

Step 2: Complete authentication and routing

Reuse login unless the user requests a login-state change. Use user_credit as the post-login readiness check. Route generation to the matching execution skill.

Step 3: Submission and terminal-state closure

Warn about credit consumption, distinguish help inspection from real submit, and follow every accepted task until success or fail.

Validation

  • Validate the installed CLI help against the reviewed contract before changing model or resolution assumptions.
  • Discover video ratio support at runtime with dreamina <subcommand> --help: four video commands expose --ratio, while multiframe2video infers it from the first image.
  • Validate local files before upload and preserve every returned submit_id.
  • Validate terminal status from query_result, not shell exit code alone.

Gotchas

  1. Help drift: release notes are not a substitute for current subcommand help.
  2. Double fact source: keep parameter matrices in the v1.4.18 contract reference.
  3. Accepted versus successful: querying is not terminal success.
  4. Paid action: warn before any real generation submit.
  5. OAuth pause: always report whether login succeeded, was reused, or failed.
  6. Troubleshooting order: capture the exact command, error, dreamina version, and relevant ~/.dreamina_cli/logs/ excerpt; redact secrets, update the CLI with authorization, then retry the same command once.

Login completion: mandatory user-visible confirmation

dreamina login / dreamina relogin prints OAuth Device Flow instructions and then waits for authorization. When the command finishes successfully, tell the user explicitly that login succeeded or the local OAuth state was reused.

  • Do not wait for the user to ask “登录好了吗”.
  • Do not stop after only sending the device code: keep the login command running, read stdout to the end, then confirm success/reuse/failure.
  • Failure must still be reported with the concrete error and the next step.

Choosing the right command

At a high level:

  • Use user_credit to check budget.
  • Use session to create, list, search, rename, or delete sessions; all generator commands accept --session=<id> and 0 is the default session.
  • Use query_result when you already have a submit_id; add --download_dir when you want the generated media saved locally.
  • Use list_task to review recent saved tasks, especially when you want to filter by status or task type.
  • Use text2image for prompt-only image generation, image2image for image-guided editing, and image_upscale for upscaling.
  • Use text2video for prompt-only video generation.
  • Use image2video when one main image is enough; if the user has multiple images for a coherent story, prefer multiframe2video.
  • Use frames2video for first-and-last-frame driven video generation.
  • Use multiframe2video for Dreamina's intelligent multi-frame flow: multiple images in, one coherent story video out.
  • Use multimodal2video for Dreamina's flagship video mode when the task needs all-around references across images, video, and audio; it supports the seedance2.0 family. If the legacy name ref2video appears, trust dreamina -h for the current command surface.

For the exact flags and supported combinations, rely on each subcommand's -h.

Model selection rule

Do not hardcode model support from this skill.

If the user specifies a model, always check the relevant subcommand help before running it:

dreamina <subcommand> -h

Use the subcommand help to confirm:

  • whether that command exposes model selection
  • whether the requested model is supported on that command
  • what other constraints apply to that model, such as duration, ratio, resolution, or whether the command supports model_version at all

Additional guidance:

  • some commands do not expose model selection at all
  • some models, especially the seedance2.0 family, can be capacity-constrained
  • if the user cares more about speed than maximum quality, do not default to seedance2.0 unless they explicitly ask for it

How to judge submit acceptance and terminal success

Do not rely on shell exit code alone.

For async generation commands, submit_id plus gen_status=querying means only that the submission was accepted. Treat the generation as terminally successful only when gen_status=success. If gen_status=fail, inspect fail_reason and reply proactively.

Follow-up pattern for async tasks

After a submit returns querying:

  1. Save the submit_id.
  2. Use query_result --submit_id=<id> for follow-up.
  3. Use list_task when you want to review saved tasks in bulk.
  4. Continue until the task reaches success or fail.

If you are running a test sweep, keep results in a machine-readable format so you can query the returned submit_id values later.

Important user-facing rules

  • Some generation commands are asynchronous; submit and query are separate steps.
  • Some models may require a one-time authorization on Dreamina Web. If the CLI returns AigcComplianceConfirmationRequired, reply proactively: ask them to complete that web-side confirmation first, then retry.
  • Do not assume that different commands support the same models, ratios, durations, or resolutions. Check each subcommand's -h before use.
  • In v1.4.18, text2video and multimodal2video accept explicit --ratio; image2video and frames2video expose the flag but reject it with seedance2.5; multiframe2video has no ratio flag. Re-discover these rules at runtime instead of generalizing across commands.

Good agent behavior

  • Relay OAuth Device Flow instructions exactly enough for the user to complete login.
  • Always close the loop when the login command finishes with a user-visible confirmation.
  • Prefer small, reviewable batches when running real generation tasks.
  • Keep a record of the command, arguments, submit_id, and final status for every paid test you run.
  • When the user cares about generation speed, do not default to the seedance2.0 family unless they explicitly ask for it or clearly prioritize output quality.
  • If you are preparing a report, separate:
    • help-only inspection
    • submit-stage validation
    • later async result follow-up

When to use (什么时候使用)

Load this skill when the user needs to select and execute a verifiable, recoverable professional workflow for the current request . First extract the goal, inputs, constraints, delivery format, and acceptance criteria from the request; the description summary is: Use when an agent needs to install, update, authenticate, troubleshoot, or operate Dreamina (即梦) CLI across account, session, image, video, task-query, and download workflows.

Rules

  • Read before write: first confirm the current state and real capabilities, then perform actions that change external state.
  • Least privilege: use only the files, tools, accounts, and network scope required to complete the current step.
  • Evidence first: when run results, resource IDs, versions, hashes, or test outputs are missing, clearly mark them as NOT_VERIFIED。
  • Idempotency first: retain request identifiers and stage status; when results are unclear, query first rather than blindly retrying.
  • Privacy and security: logs, examples, receipts, and error messages must not contain tokens, cookies, keys, or personally sensitive data.

Workflow

Step 1: Clarify intent

Confirm whether this skill matches the goal; if it is merely an adjacent need, hand it off to a more precise skill.

Step 2: Perform prechecks

Confirm the goal, inputs, constraints, available tools, success criteria, and failure boundaries; if any key condition is unknown, stop at the read-only stage.

Step 3: Form a plan

List the tools to be called, the objects that will be changed, the success criteria, and the safe exit method after failure.

Step 4: Execute actions

Execute in minimal sufficient steps, and record evidence at key state changes; every external call retains an associable state or receipt.

Step 5: Verify delivery

Output the results, verification evidence, unfinished items, risks, and clear next steps, and state facts, inferences, and unverified items separately.

Validation checklist

  • The skill trigger conditions match the user's intent, and no adjacent task was misrouted to this skill.
  • The inputs, target objects, versions, and output locations are all clear, and no guessed values were used in place of required values.
  • All write, paid, publish or irreversible actions are within the scope of user authorization.
  • The result has been verified with independent checks; only "command succeeded" or "file exists" does not count as complete acceptance.
  • The output includes actual evidence, failed/skipped items, remaining risks and executable next steps.

Gotchas

  1. Treating the plan as the result: A document or prompt is not equal to actual execution; the actual execution level must be indicated.
  2. Error retry: A timeout or lost response may already have produced remote state; query first, then decide whether to retry.
  3. Implicit scope expansion: Batch, full-scale, publish, overwrite and paid actions are not a natural extension of ordinary reads and writes.
  4. Version drift: When referencing external resources, record the version, tag or commit; do not treat a mutable branch as release evidence.
  5. Stale evidence: Caches, old screenshots and historical tests cannot prove the current environment; refresh key evidence before delivery.

Non-applicability and boundaries

Do not exceed the scope given by the user; write, paid, publish and irreversible actions require explicit authorization. If the request requires another skill, do not copy its body text; hand off by skill name and preserve the current task context.

Progressive disclosure

  • When input/output, state and authorization points need to be determined, read references/workflow-contract.md。
  • When pre-delivery self-check is needed, read references/validation-checklist.md。
  • When encountering timeout, partial success or recovery scenarios, read references/error-recovery.md。
  • For first run, privilege escalation refusal and failure recovery, refer to examples/happy-path.md、examples/boundary-refusal.md、examples/failure-recovery.md。