shengsheng90/dsh-taskboard/manage-taskboard
Claude Code and agent skill by shengsheng90/dsh-taskboard
- community
Checking that the SKILL.md file is still in the repository.
Manage work in the native DeepSeek Harness Taskboard with exact task ids and optimistic versions. Use when an Agent must inspect project work, claim an eligible todo, record progress or blockers, verify an implementation, submit it for human review, or release its own claim; also use when a human asks how to accept, return, archive, or automate Taskboard work through the native UI or dsh-taskboard JSON CLI.
Install
npx vskill@latest install shengsheng90/dsh-taskboard/manage-taskboardSKILL.md
Manage Taskboard
Use the in-process taskboard_* tools for Agent work. Use dsh-taskboard only for human-operated scripts and interoperability; do not make the model shell out when a native tool exists.
Execute one task
- Call
taskboard_listwith the exact project id. Prefer an eligibletodo; do not selectbacklog, archived, dependency-blocked, or already claimed work. - Call
taskboard_getimmediately before claiming. Preserve the opaque task id and currentversionexactly; never derive an id from a display key such asDSH-42. - Call
taskboard_claimwith that id and version. Treat a stale-version or claim conflict as a signal to reread and reconsider, not to retry blindly. - Read the full description, comments, relations, dependency state, development context, and attachment references before changing files. Work only in the task's declared workspace, branch, or worktree.
- Complete the work and run relevant verification. If requirements change, call
taskboard_getagain before continuing. - Record the final result with
taskboard_commentortaskboard_submit_review, always using the version returned by the latest read or write. Never modify the task description. - Call
taskboard_submit_reviewwith a concise result comment and concrete verification evidence. This moves owned work toin_review; it never marks workdone.
Keep every write version-linear: after any successful comment, relation, block, or other mutation, use its returned version for the next write. If another actor wins the race, reread the task and reconcile instead of overwriting their change.
Handle exceptional outcomes
- Call
taskboard_blockonly on the in-progress task you hold the claim for, and only with a concrete reason when work cannot proceed. Include the missing dependency, decision, permission, or external condition. Atodoyou have not claimed is not yours to block: report the obstacle in a comment and leave the column to a human. - Call
taskboard_release_claimwhen intentionally abandoning owned work. Explain what remains and leave useful progress in a comment first when possible. - Use
taskboard_relateonly after reading both tasks. Keep relations within one project and do not create parent cycles. - Never call or emulate acceptance. Only a human may accept
in_reviewasdone, return it for rework, approve backlog work, archive it, or permanently delete it.
Guide human review
Ask the human to open the native Taskboard task detail, inspect the result comment and verification evidence, then choose Accept or Return for rework. For headless human automation, use dsh-taskboard task accept --task <opaque-id> --version <exact-version> or task return ... --comment .... Every successful CLI response is versioned JSON; exit codes distinguish usage, unavailable service, API errors, and optimistic conflicts.
Do not bypass service policy with direct SQLite access, browser-side file access, generic status mutation, or prompt-only assumptions. The SQLite Provider and Host service are authoritative.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/manage-taskboard/SKILL.md
Popularity
- Repository stars
- 330
- Repository forks
- 8
- 7-day trend score
- 34
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
[](https://verified-skill.com/skills/shengsheng90/dsh-taskboard/manage-taskboard)