xiangruiai/xiangrui-toolkit/manage-taskboard

Claude Code and agent skill by xiangruiai/xiangrui-toolkit

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Manage Xiangrui Taskboard projects, issues, relations, and comments. Use when Codex needs to record a requirement, inspect project work, create or update issues, add progress notes, begin execution, hand off review, or coordinate concurrent task updates.

Install

npx vskill@latest install xiangruiai/xiangrui-toolkit/manage-taskboard

SKILL.md

Manage Taskboard

Use the plugin-bundled taskctl wrapper for every project, issue, and comment operation. Resolve the plugin root from this file (../..) and invoke node <plugin-root>/scripts/taskctl.mjs .... Read references/cli.md before choosing a command or option. When that reference shows taskctl, substitute the bundled wrapper command.

Workflow

  1. Search for an existing issue before creating one. Use context current, then list the project issues and compare their identifiers, titles, descriptions, and status.
  • If an issue already tracks the same requirement, append the new requirement or acceptance detail to that issue without discarding its existing scope.
  • If the work depends on, blocks, is blocked by, or is closely related to another issue, add the matching issue relation.
  • Use a parent/sub-issue relation when one requirement is a contained part of a larger issue. A child has one parent; a parent may have many sub-issues.
  • Create a new issue only when no existing issue reasonably tracks the requirement.
  • Do not create, append, or relate a tiny or trivial request that does not benefit from durable tracking.
  1. Before executing an issue, read the latest issue content and all comments. Treat comments as part of the current requirements, especially when completed work has been returned for changes.
  • In a description or comment, ![alt](/api/attachments/<id>/content) marks an inline image at that exact position in the text.
  • When understanding that image is necessary, use attachment download to save it locally, then inspect the saved file with an available image-viewing tool.
  1. Before implementation, honor the issue's execution metadata.
  • For a bound worktree, run commands and edit files only in its exact path.
  • For a bound branch, prefer an existing worktree checked out at that branch; otherwise verify the current workspace is already on the bound branch before editing. Never silently execute against a different branch. If the bound context cannot be used safely, add a comment with the exact unblock condition and move the issue to blocked with both --blocked-reason and --unblock-action.
  • Preserve dueDate during claims and handoffs. It affects automatic pickup order but does not override status, dependency, or review rules.
  1. Create or update issues with the CLI; consume its JSON output.

Issues created through taskctl are assigned to Codex Agent by default. Later CLI updates do not change the assignee.

  1. Let taskctl attribute every issue, relation, or comment mutation to the current Codex conversation through CODEX_THREAD_ID. Outside Codex, pass the exact conversation id with --thread-id.
  2. To claim a todo issue, move it to in_progress with --if-version from the latest read before starting implementation. If this claim reports a version conflict or a new read shows that its status changed, skip the issue and do not implement it.
  3. Include --if-version <version> on every concurrent update, using the version returned by the latest read.
  4. Before requesting review, verify the requested work and acceptance criteria.
  5. An issue may remain in_progress only while Codex is actively working on it. Before a normal execution exits, choose the truthful outcome and add a comment explaining it:
  • Completed and self-verified: move to in_review after the result comment.
  • Temporarily interrupted but safe to retry without new external input: record the exact resume point and move back to todo.
  • Waiting for user input, missing information, or an unfinished dependency: move to blocked with --blocked-reason containing only the confirmed fact and --unblock-action containing one explicit action that will let work resume.
  • Never infer, summarize, or invent a blocking reason from status, comments, age, or incomplete context. Use explicit user input or a verified tool/runtime error as the source.
  • Never move to canceled unless the user explicitly decides the work should not continue.

If a process terminates before cleanup, a later automation may recover the issue only after its threadId is confirmed failed or interrupted through Codex task/thread state. Never infer termination from age alone and never reset an active run.

  1. After implementation and self-verification, add a comment summarizing the key changes, verification, result, and remaining risks; then move the issue to in_review. Never move it directly to done.
  • When the handoff prompt names a source comment, pass that comment id with comment add --reply-to so the result visibly links back to the feedback being addressed.
  1. Move an issue from in_review to done only when the user explicitly confirms acceptance or explicitly asks to mark it complete. Codex self-verification alone is not sufficient.
  2. Move work that cannot continue to blocked, always including the exact structured reason and unblock action. Move work that will not continue to canceled.

For version conflicts outside the initial claim, read the issue again, reconcile the newer state, and retry with its current version.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
55
Repository forks
43
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/xiangruiai/xiangrui-toolkit/manage-taskboard/badge)](https://verified-skill.com/skills/xiangruiai/xiangrui-toolkit/manage-taskboard)

Something wrong with this skill or its page? Tell us.