菜单
Regenerate all _index.md dashboard files (weekly, people, projects, decisions, incidents, hiring, strategy) from current content. Use when the user asks to rebuild, refresh, or regenerate the dashboards/indexes. No args needed.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Create a new GitHub issue (optionally added to a GitHub Project board) following the project's standard format — gathers the details, shows a draft for confirmation, then creates it via the `gh` CLI. Use when the user wants to file, open, or create a GitHub issue. Optionally pass the repo and/or issue title as args.
Explore a remote GitHub repository read-only via the `gh` CLI — search code, read a file at a ref, view repo structure, trace the history/PRs touching a path, and blame lines. Use when investigating an incident or working an assignment against code you don't have checked out, across one or more repos. Pass the repo(s) (`owner/repo`) and what you're looking for as args.
Fetch a GitHub pull request as a structured bundle (metadata, diff, changed files with stats, commits, CI/check status, and existing review comments) via the `gh` CLI. Transport-only — it retrieves, it does not judge. Used by the tool-agnostic `pr-review` skill; you normally won't invoke it directly. Pass a PR number, URL, or branch (and optionally `owner/repo`) as args.
Sync a GitHub Project board into an emOS report grouped by status. Use when the user wants an up-to-date snapshot of a GitHub Projects (v2) board pulled into 10-reports/. Optionally pass a project number/URL, an owner, or a status filter as args.
Do a first-line review pass on a pull request — summarize the changes, flag obvious issues, highlight architectural/dependency changes, and escalate what needs human attention. It hands you structured input to act on; it does NOT auto-approve or post to the code host. Use when the user wants a PR reviewed, triaged, or summarized before they dig in. Pass a PR number, URL, or branch as args.
Import any document by URL into emOS. Detects the transport (Confluence, Google Docs, GitHub, or a generic web page), delegates to the medium-specific provider skill when its capability is configured, classifies the content, files it, and pulls any action items assigned to the owner into the current week's plan. Use when the user gives a URL and wants its content brought into the vault. Pass the URL (and optionally a hint about the doc type) as args.
| name | rebuild |
| description | Regenerate all _index.md dashboard files (weekly, people, projects, decisions, incidents, hiring, strategy) from current content. Use when the user asks to rebuild, refresh, or regenerate the dashboards/indexes. No args needed. |
Rebuild all dashboard index files from current content.
Read owner.role from .agents/config.yaml (default em) and open the role pack at
.agents/roles/{role}.md for its "Sections in focus / dormant" note. Build every
dashboard whose folder has content (they degrade to a minimal placeholder when empty);
order and lead with the role's in-focus sections. A role never removes a dashboard — it
only changes emphasis.
For each of the following dashboards, scan the relevant content files and generate a fresh _index.md:
Scan all 01-weekly/YYYY-Www.md files. Build a table with: Week, Status, completion rate (count [x] vs [ ]), and any #blocked items. Highlight the current active week.
Scan all 02-people/profiles/*.md. Build a roster table: Name, Role, Relation, Level, Status. Show the Skill/Will rating only for people with relation: report (blank otherwise). Note anyone with status other than active. Also list 1-on-1 cadence (count files per person in one-on-ones/).
Scan 03-projects/*/ (each subfolder except archive/), read the hub {slug}.md file inside. Group by status (Active · Proposed · Paused · Completed · Cancelled). For each project list: name (linked to the hub doc), priority, quarter, and a Jira epic link if jira_epic is set. Then scan 03-projects/archive/*/ into a Completed/Archive table. Do not touch the <!-- rollup --> blocks inside hub docs — those are owned by project-sync.
Scan 04-decisions/*.md. Build a table: ADR ID, Title, Status, Date, Deciders. Group by status.
Scan 05-incidents/*.md. Build a table: Date, Title, Severity, Status, Owner. Sort by date descending. Flag any not yet reviewed.
Scan 06-hiring/positions/*.md and interviews/*.md. Build: open positions table and candidates-in-flight table with stage and recommendation.
Scan 07-strategy/okrs/*.md and roadmaps/*.md. Summarize active OKRs with status and key results count. List roadmap files.
After regenerating each file, confirm the path and a one-line summary of what changed. If a directory has no content files, write a minimal placeholder.
