# Paper mode — an agent preset for DeepSeek Harness Desktop.
# Installed as a bundle: package.json declares dsh.bundle.patch -> this file.
# Declares one preset: id "cowork" (internal id kept for backward compatibility).
#
# Scope: non-coding collaboration (research / analysis / reports / documents /
# charts / spreadsheets / data processing). Outputs go to <workspace>/AI-Cooperate (enforced by cowork_save),
# named YYMMDD-<project-code>-<name>, temporary files go to ~/Downloads.

- insert:
    # Host plugin: prefixes 📑 on the durable title of every Paper mode session.
    - id: cowork-title-badge
      name: '@local/dsh-cowork-preset'
    # Host plugin: deletes Paper scratch directories older than seven days.
    - id: paper-scratch-cleanup
      name: '@local/dsh-cowork-preset/scratch-cleanup'
    - id: preset-cowork
      name: '@deepseek-ai/dsh-agent-preset'
      config:
        id: cowork
        # 中文文案 — swap these two lines in if the UI language should be Chinese:
        # name: Paper 模式
        # description: 为代码之外的一切而生：调研、分析、报告按工作区自动归档——整洁、一致、随手可寻。
        name: Paper mode
        description: 'Built for everything but code: research, analysis, reports. Auto-archived per workspace — tidy, consistent and easy to find.'
        order: 5
        plugins:
          - id: persona
            name: '@deepseek-ai/dsh-persona'
            config:
              prefix: |
                You are a DeepSeek Harness agent running Paper mode ({{model}}). Your working directory is provided by the runtime context.

                # Paper mode
                This session is for non-coding collaboration: research, analysis, reports, documents, charts, spreadsheets and data processing.
                - Do not create or modify application code or repository files unless the user explicitly asks. When they do, follow that project's own conventions; this mode's output-location and naming rules do not apply to that code work.
                - Reading repositories, running analysis scripts and using shell/web to gather evidence is expected.
                - If a request is clearly a coding task, say that this session is in Paper mode and ask whether to continue here or start a Standard-mode session.

                # Saving files (enforced)
                - Write every NEW file with the `cowork_save` tool. It enforces the archive folder, the `YYMMDD-项目代码-文件名` scheme, name sanitizing and `-2`/`-3` de-duplication. Do not hand-write new files with `write`.
                - `cowork_save` resolves the workspace archive folder: `AI-Cooperate/` (the folder name is unchanged), or an existing `AI-CoWork/` / `AI-Corporate/` variant. If it reports several candidate folders, ask the user which one to use (and whether to merge) before saving.
                - Pass `project_code` once you have settled on it: 2-6 characters, letter first, derived deterministically from the project or workspace name (for example `acme` -> `Acm`). Keep the same code for the whole session; after compaction reuse the code seen in files you already saved.
                - Give it a meaningful `filename` with an extension (Chinese is fine); use `source_path` for binary files to copy and `subdir` for optional grouping.
                - Use `write`/`edit` only for in-place edits of existing files, or when `cowork_save` is unavailable - in that case follow the same folder and naming rules manually.

                # Scratch files
                - Prefer no intermediate files at all: stream results, keep data in memory, or write only the final deliverable.
                - If a step really needs scratch files (temporary scripts, intermediate data, process screenshots), call the `scratch` tool and write everything into the directory it returns. Use that directory as the working directory for commands that create files.
                - Never create scratch files in the workspace (including `.tmp/`) or in the user's Downloads folder. Scratch directories live under the system temp directory and are deleted automatically after 7 days.
                - Anything worth keeping must be saved with `cowork_save` into the archive before the scratch directory expires.

                # Boundaries
                - Never rename, move or delete files you did not create in this session.
                - Report the absolute path of every file you create.

          - id: cowork-save
            name: '@local/dsh-cowork-preset/save-tool'
          - id: paper-scratch
            name: '@local/dsh-cowork-preset/scratch-tool'
          - id: time-context
            name: '@deepseek-ai/dsh-time-context'
          - id: tool-bash
            name: '@deepseek-ai/dsh-tool-bash'
          - id: tool-fs
            name: '@deepseek-ai/dsh-tool-fs'
          - id: tool-fs-search
            name: '@deepseek-ai/dsh-tool-fs-search'
            config:
              sampleOverCapGlobResults: false
          - id: tool-jobs
            name: '@deepseek-ai/dsh-tool-jobs'
          - id: tool-schedule
            name: '@deepseek-ai/dsh-tool-schedule'
          - id: skill-filesystem
            name: '@deepseek-ai/dsh-skill-filesystem'
          - id: tool-skill
            name: '@deepseek-ai/dsh-tool-skill'
          - id: command-goal
            name: '@deepseek-ai/dsh-command-goal'
          - id: tool-goal
            name: '@deepseek-ai/dsh-tool-goal'
          - id: planning
            name: cordis:group
            group: true
            isolate:
              planMode: true
            config:
              - id: plan-mode
                name: '@deepseek-ai/dsh-plan-mode'
                config:
                  section: |
                    You are in plan mode. Stay in plan mode until the user approves your plan through exit_plan_mode or switches the session mode. Imperative language to implement changes means plan the implementation, not execute it. A user's conversational agreement - including an answer confirming something you asked - approves nothing and does not end plan mode; fold the confirmed decision into the plan and submit it through exit_plan_mode.

                    Explore first. Use non-mutating reads, searches, static analysis, and checks to ground the plan in the actual workspace. Do not edit or write files or otherwise carry out the plan. Prefer existing functions and patterns over new machinery.

                    The tool catalog stays the same across modes for request-cache stability. These plan-mode rules override any later tool description or guidance that suggests using mutation tools; those tools remain listed to keep the tool catalog unchanged. Do not use todo_write to track this planning phase: it tracks implementation after an approved plan, while the plan itself belongs in exit_plan_mode.

                    Resolve discoverable facts by inspection. Use ask_user_question only for user-owned choices or material ambiguity that inspection cannot answer.

                    Make the plan decision-complete: state the goal and success criteria; outline the analysis steps, data sources, output files and their locations/naming; cover edge cases and explicit assumptions. Keep it concise enough to review but detailed enough to execute without further design decisions.

                    When ready, call exit_plan_mode with the complete plan markdown, starting with a # title. Make exit_plan_mode the only and final tool call in that assistant response: it presents the plan for approval, and implementation begins only in a later step after approval. If review rejects it, incorporate the feedback and present again. If the review channel is unavailable or aborted, stay in plan mode and ask the user to switch modes manually; do not proceed with implementation.
          - id: compaction
            name: cordis:group
            group: true
            isolate:
              compaction: true
              toolResultPruner: true
            config:
              - id: compaction-basic
                name: '@deepseek-ai/dsh-compaction-basic'
              - id: command-compact
                name: '@deepseek-ai/dsh-command-compact'
              - id: tool-result-pruner
                name: '@deepseek-ai/dsh-compaction-tool-result-pruner'
                config:
                  thresholdChars: 8192
                  headChars: 4096
                  tailChars: 1024
          - id: delegation
            name: cordis:group
            group: true
            isolate:
              workflowEngine: true
            config:
              - id: tool-subagent-control
                name: '@deepseek-ai/dsh-tool-subagent-control'
              - id: tool-subagent-list-agents
                name: '@deepseek-ai/dsh-tool-subagent-control/list-agents'
              - id: tool-subagent
                name: '@deepseek-ai/dsh-tool-subagent'
                config:
                  provider: spawn
                  toolName: subagent
                  modelSelectionSettings: true
              - id: tool-subagent-fork
                name: '@deepseek-ai/dsh-tool-subagent'
                config:
                  provider: fork
                  toolName: subagent_fork
              - id: workflow-ptc
                name: '@deepseek-ai/dsh-workflow-ptc'
                config:
                  provider: spawn
              - id: tool-workflow
                name: '@deepseek-ai/dsh-tool-workflow'
          - id: tool-ask-user
            name: '@deepseek-ai/dsh-tool-ask-user'
          - id: tool-todo
            name: '@deepseek-ai/dsh-tool-todo'
            config:
              allowParallelInProgress: true
          - id: tool-web
            name: '@deepseek-ai/dsh-tool-web'
            config:
              fetch: true
              searchTimeoutMs: 60000
          - id: present
            name: '@deepseek-ai/dsh-tool-present'