agentsclimarketplace

Build work memory

Skill stardustai/stardust-skills/skills/build-work-memory

面向钉钉与叮当 OKR 工作流的本地 Agent Skills

Install
npx -y skills add stardustai/stardust-skills --skill build-work-memory

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Guided first-run workflow for importing a user's existing work context into Friday Memory from local files, DingTalk documents/wiki, DingTalk AI minutes, and selected DingTalk data through dws. Use when the user wants to initialize Memory, connect work context, import DingTalk knowledge, seed Memory from documents or meetings, or prepare Memory for daily work.

SKILL.md

5.8 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it

Build Work Memory

Import the user's existing work context into Friday Memory.

Run with Codex, Friday Memory MCP, and the dws DingTalk Workspace CLI. Do not require the memory-connector repository, Neo4j, Friday admin UI, local backend services, or project-local scripts. Use local file access only when the user explicitly chooses local documents.

Preflight

Before asking import questions, verify:

  • Friday Memory MCP is available and exposes document_upload.
  • dws is installed and authenticated.

If Friday Memory MCP is missing, tell the user to open https://friday.stardust.ai, sign in, generate the Memory MCP / Agent install command, and send that command to Codex. Do not invent install commands or tokens.

If dws is missing, ask the user for the approved installation source or internal installer. Do not invent a download URL. After installation, verify with:

dws --help
dws profile list --format json

Experience Contract

Do not ask the user for a project or topic first. This skill is for first-time work-context initialization.

Do not default-scan anything. Only import ranges the user explicitly describes.

Do not require per-item confirmation. Collect scope, summarize the planned read and upload range, and ask once for explicit confirmation before reading private content or uploading to Memory.

Do not expose graph_id, user_id, document_upload, ingest_mode, entity extraction, relation extraction, or internal queue details to the user.

Do not run recall validation. Graph extraction is asynchronous and may take time.

Opening Message

Start with:

我会帮你把已有工作上下文导入 Memory。流程分 4 步：

1. 本地文档
2. 钉钉知识库 / 钉钉文档
3. 钉钉 AI 听记 / 会议纪要
4. 其他钉钉数据

每一步我都会问你想导入的范围。收集完范围后，我会在正式读取和上传前汇总一次；你确认后我再执行。不会默认扫描你的全部资料。

Scope Collection

Ask in order:

  1. Local documents: paths, folders, file types, keywords, or time range.
  2. DingTalk documents/wiki: workspace, folder, document links, keywords, owner, or time range.
  3. DingTalk AI minutes: meeting keywords, time range, participants, links, or IDs.
  4. Other DingTalk data: ask whether any specific category and range should be included.

For other DingTalk data, present the supported categories:

  • AI search / people lookup
  • AI tables
  • Attendance
  • Calendar
  • Chat / IM
  • Contacts / departments
  • DING messages
  • DingTalk drive
  • Live sessions
  • Mail
  • OA approvals
  • Reports / daily or weekly logs
  • Online sheets
  • Todos
  • Wiki spaces and members
  • Developer docs / error-code docs
  • PAT / authorization diagnostics, excluding tokens or secrets

If the user gives an overly broad range, ask them to narrow it before reading. Examples: all company wiki, all meetings, all chats, all drive files, all data from a full year.

Confirmation

After collecting scope, summarize:

我将按以下范围读取资料并导入 Memory：

- 本地文档：...
- 钉钉知识库 / 文档：...
- 钉钉 AI 听记：...
- 其他钉钉数据：...

请确认是否开始读取并上传这些范围内的内容。

Proceed only after the user confirms.

DingTalk Rules

Use only dws for DingTalk data. Use --format json.

If command syntax is uncertain, inspect it first with dws <service> --help or dws schema <path>.

Route sources through:

  • dws doc for DingTalk documents
  • dws wiki for wiki spaces/containers
  • dws minutes for AI minutes
  • dws chat, calendar, mail, drive, report, todo, sheet, aitable, contact, aisearch for selected optional data

Memory Ingest Format

Convert every imported source into Markdown before upload. Add provenance at the top:

# <title>

source_type: <local_file | dingtalk_doc | dingtalk_wiki | dingtalk_minutes | dingtalk_chat | dingtalk_calendar | dingtalk_mail | dingtalk_drive | dingtalk_report | dingtalk_todo | dingtalk_sheet | dingtalk_aitable | dingtalk_contact | other>
source_id: <id if available>
source_url: <url if available>
source_path: <folder/wiki/path if available>
source_time: <original created/updated/meeting time if available>
participants: <participants if available>
imported_from: <local | dws doc | dws wiki | dws minutes | ...>

---

<content>

Use original source time for uploaded_at when available. Add a concise summary and background when possible.

Split long sources by natural boundaries: headings, wiki sections, agenda items, speaker/time segments, dates, table views, or logical chapters. Keep provenance in every split part.

Upload Rules

For each prepared document, call Friday Memory MCP document_upload with:

  • filename
  • content_base64
  • mime_type
  • uploaded_at
  • source_kind: system_auto_fetch for DingTalk, user_upload for local files
  • summary
  • background
  • ingest_mode: graph
  • wait_for_processing: false

Never pass user_id, graph_id, or graph_ids.

Treat upload success as "submitted to Memory processing," not "fully extracted."

Final Response

End with:

已提交到 Memory：

- 本地文档：N 份
- 钉钉知识库 / 文档：N 份
- 钉钉 AI 听记：N 份
- 其他钉钉数据：N 份

失败：N 项
后台处理中：N 项

这些资料会继续在后台抽取实体和关系。大型资料可能需要一段时间后才能完整召回。

List failed items with short reasons and the next retry action.

Gives 0 of the 12 instructions most memory context skills give in ~1.4k tokens

Counted across 674 of the 847 authors here whose files we hold, read 2026-08-06

  • inform the user when setup is completein 21 of 674, across 6 files
  • confirm the draft with the user before writingin 21 of 674, across 6 files
  • update the agent skills block in place if it existsin 21 of 674, across 6 files
  • present findings to the userin 20 of 674, across 5 files
  • write the three docs files from seed templatesin 20 of 674, across 5 files
  • ask the user about each decision one at a timein 19 of 674, across 4 files
  • edit CLAUDE.md if it existsin 18 of 674, across 3 files
  • explore current repo statein 18 of 674, across 3 files
  • do not overwrite user edits to surrounding sectionsin 18 of 674, across 3 files
  • back up the original file before overwritingin 16 of 674, across 8 files
  • keep the memory index under 200 linesin 15 of 674
  • Provide actionable steps and verificationin 13 of 674, across 2 files

Said here and by no other author read

  • verify friday memory mcp exposes document_upload
  • verify dws is installed and authenticated
  • summarize the planned read and upload range
  • ask once for explicit confirmation before reading
  • proceed only after user confirmation
  • use only dws for dingtalk data

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.