/alibabacloud-dts-doc-parse

name: alibabacloud-dts-doc-parse
description: Use Alibaba Cloud DTS-AI for end-to-end parsing of local PDF, DOCX, PPTX, and image files such as JPEG, JPG, and PNG into Markdown files. Use when the user mentions DTS-AI document parsing or wants to parse, convert, or extract local PDF, Word, PowerPoint, JPEG, JPG, or PNG files into Markdown.

Document Parsing

Capabilities and supported formats

Use `dtscli doc parse` to parse supported local files into Markdown and deliver the final files to the user. Cloud jobs and their recovery mechanism are internal implementation details; in the normal workflow, never ask the user to inspect, select, or resume a job.

Parse readable, non-empty local files with these extensions: `pdf`, `docx`, `pptx`, `png`, `jpg`, and `jpeg`.

Cloud processing and safety boundaries

• Start parsing only when the user explicitly requests that DTS-AI or this Skill parse specified local files, or explicitly confirms after being informed that the workflow requires cloud processing. Authorization covers the upload and cloud processing required for the current request. Never upload a file by itself or process a file the user did not specify.
• Never print, log, or expose the API key or temporary OSS credentials.
• Read credentials only through `dtscli`. Read [DTS-AI authentication](references/dts-ai-authentication.md) when the API key is missing, setup is required, or authentication fails.
• Save the Markdown written by `dtscli`. The CLI already converts HTML tables to GFM and strips nested HTML bold; do not rewrite the file again. After a timeout, interruption, save failure, or indeterminate result state, read [parser job execution and result handling](references/parser-job-execution.md).

Runtime and authentication setup

Require `dtscli` on PATH. macOS, Linux, and Windows are supported. When the shell reports `dtscli: command not found`, read [platform installation and command execution](references/runtime-platforms.md) to select the installer, paths, and shell syntax for the actual host.

In the current user session, run `dtscli version --json` only the first time this Skill is used. Confirm that capabilities contains `skills.ensure`, then run:

dtscli skills ensure --agent <agent>

Then run `dtscli ai check`. Later steps and sub-agents in the same session run business commands directly. Continue when `ready=true`, including `action=busy`. Stop and reload Skill files when `reload_required=true`. When `ready=false` without reload, read [automatic Skills updates](references/skills-updates.md).

`--agent` names the current host. `skills ensure` uses it to select the install directory; `doc parse` / `doc resume` send it as the DTS-AI call source. The values are the same.

| Current host | `--agent` | | --- | --- | | Codex | `codex` | | Claude / Claude Code | `claude` | | Cursor | `cursor` | | ZCode | `zcode` | | Qoder | `qoder` | | QoderWork | `qoderwork` | | Qwen Code | `qwencode` | | QwenWorkCN | `qwenworkcn` | | OpenCode | `opencode` | | Kimi | `kimi` | | Pi | `pi` |

When the current host can be identified, always supply `--agent`. The value must match `[a-z][a-z0-9_-]{0,31}`. An unlisted host uses its own stable lowercase identifier. If the host cannot be identified at all or the CLI rejects the value, read [calling agent source](references/agent-source.md). Replace `<agent>` in the examples below with the current host identifier.

Standard output carries only the final JSON result object and step-by-step progress goes to standard error, so standard output can be piped straight into a JSON parser. On failure, standard output holds an object containing `error_class` and the exit code is non-zero.

If the CLI is missing, obtain installation authorization and follow [platform installation and command execution](references/runtime-platforms.md).

Then run `dtscli version --json` to verify the version and continue the local authentication check. Never fall back to another way of calling DTS-AI or ask for an API key in chat.

If the key is absent, read [DTS-AI authentication](references/dts-ai-authentication.md) before offering:

dtscli ai setup --language <en|zh>

Request identity

Before any cloud API invocation, read this skill's [package identity manifest](references/manifest.json). Use only its top-level non-empty string `name` as `{skill-name}` and only its top-level non-empty string `version` as `{skill-version}`. The `name` must match this Skill's frontmatter `name` and the skill directory name. If missing, invalid, or inconsistent, STOP. NEVER invent, guess, or reuse a name or version from another skill. On a skill switch or return, reread that skill's package identity manifest. Read [request identity](references/request-identity.md) before constructing the request identifier.

Before this skill's first cloud API invocation in a conversation, generate a fresh random 32-character lowercase hexadecimal session ID. Reuse that session ID for this skill throughout the conversation; each skill MUST use a distinct session ID. NEVER copy one from documentation, examples, another skill, or a previous conversation, and NEVER send the literal `{session-id}` placeholder.

Attach `--user-agent` only to `dtscli` commands that call an OpenAPI and accept `--user-agent`. Exact user-agent with skill version propagation across CLI/SDK/Terraform:

--user-agent "AlibabaCloud-Agent-Skills/alibabacloud-dts-doc-parse/{session-id} skill-version/{skill-version}"

Output files and internal recovery

When `--output` is omitted, write `<input-name>.md` beside the input file. If that name is occupied by a different document, the command appends the first eight characters of the input SHA-256 as `<input-name>-<sha256-prefix>.md`. Recovery records stay in `.aliyun-dts/doc-parse/` under the current user's home directory. If the input directory is not writable, the command falls back to that home directory. Another output directory must a