菜单
Mark a context-task-planning task done after verification and summarize any blockers if completion is refused. Use this whenever the user wants to finish a task, mark tracked work done, close the current task after verification, or explicitly invokes /task-done.
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。
Task-scoped context engineering for complex multi-step work. Use when a task needs clarification, phased execution, durable file-based state, recovery after context loss, or optional sub-agent delegation.
Task-scoped context engineering for complex multi-step work. Use when a task needs clarification, phased execution, durable file-backed state, recovery after context loss, or optional sub-agent delegation.
Inspect the current context-task-planning task in this workspace and summarize the resolved task, status, and next action. Use this whenever the user asks what task is active, what the next step is, wants a task summary, or explicitly invokes /task-current.
Return heuristic route evidence for whether a new request still fits the current context-task-planning task and summarize the routing recommendation. Use this whenever the user asks whether a new ask belongs in the same task, wants a drift check, is about to switch scope, or explicitly invokes /task-drift.
Initialize a new context-task-planning task for this workspace from a confirmed task title and final task slug. Use this whenever the user wants to start tracking a new task, create a planning lane, initialize task files, or explicitly invokes /task-init.
List the existing context-task-planning tasks in this workspace and summarize the active pointer plus the most relevant tasks. Use this whenever the user asks what tasks exist, what they can resume, wants a task list, or explicitly invokes /task-list.
基于 SOC 职业分类
| name | task-done |
| description | Mark a context-task-planning task done after verification and summarize any blockers if completion is refused. Use this whenever the user wants to finish a task, mark tracked work done, close the current task after verification, or explicitly invokes /task-done. |
| allowed-tools | Bash |
Mark a context-task-planning task done after verification.
context-task-planning skill to be installed alongside this entry skill.${CLAUDE_SKILL_DIR}/../../skill/scripts/done-task.sh for Claude plugin installs, ${COCO_PLUGIN_ROOT}/skill/scripts/done-task.sh for TraeCLI/Coco plugin installs, $HOME/.codex/plugins/context-task-planning/skill/scripts/done-task.sh for Codex plugin installs, then standalone skill paths under $HOME/.claude/skills/, $HOME/.codex/skills/, $HOME/.config/opencode/skills/, and finally repo-local skill/scripts/done-task.sh.done-task.sh requires state.verify_commands and matching successful rows in progress.md under ## Verification Log. If the script reports missing verification evidence, blockers, active delegates, or blocked phases, stop and summarize the issue instead of trying to force completion.Run the matching command for whether a slug argument was provided.
core=""
for candidate in \
"${CLAUDE_SKILL_DIR:-}/../../skill/scripts/done-task.sh" \
"${COCO_PLUGIN_ROOT:-}/skill/scripts/done-task.sh" \
"$HOME/.codex/plugins/context-task-planning/skill/scripts/done-task.sh" \
"$HOME/.claude/skills/context-task-planning/scripts/done-task.sh" \
"$HOME/.codex/skills/context-task-planning/scripts/done-task.sh" \
"$HOME/.config/opencode/skills/context-task-planning/scripts/done-task.sh" \
"skill/scripts/done-task.sh"; do
[ -f "$candidate" ] && core="$candidate" && break
done
[ -n "$core" ] || { echo "context-task-planning done-task.sh not found" >&2; exit 1; }
# No slug argument:
sh "$core"
# With a slug argument:
sh "$core" "<slug>"