dev-create-tasks

name	dev-create-tasks
description	Create or refine durable, implementation-ready Taskwarrior development tasks without claiming or executing them. Use for filing, queuing, or decomposing development work, including follow-up findings.

Dev Create Tasks

Create pending, unclaimed work through scripts/dct-create.sh, then stop. Let CREATE_SKILL be this skill's absolute directory and LOOP_SKILL its sibling dev-loop; keep the target Git repository as the current directory. The helper derives and validates repository identity. Use --help for its argument contract.

Boundary

• Mutate only Taskwarrior. Do not claim, implement, test, or create branches.
• Keep tasks reconstructable without conversation history or agent attribution.
• Run task rc.confirmation=no sync before inspection and after the complete batch. Accept only the helper's explicit unconfigured-sync result.
• Treat AGENT_PID and AGENT_NOTIFY as controller-owned lifecycle values: inherit them verbatim; never assign, overwrite, or unset them.
• End every run by calling dl-finish.sh; see Finish.

Design the task set

Default to one task per coherent outcome. Split when work needs independent bases or acceptance decisions, or cannot fit in one isolated worktree. Give every task an outcome-oriented description and observable acceptance:.

Use both --depends <producer> and --input <producer-branch> when code ancestry matters. Put end-to-end acceptance on the final task in a stack. Use --small only for narrow, mechanical, fully specified work; never assign +LARGE during initial creation. Use --plan for a decomposition task whose deliverable is finer-grained tasks rather than code: attach its plan artifact via the implement skill's dl_plan_put, and leave it for the sibling dev-decompose-task skill — work queues never claim +PLAN tasks.

For an existing goal, run the sibling controller's read-only state helper:

"$LOOP_SKILL/scripts/dl-loop-state.sh" --goal '<goal-slug>'

Reuse its loop ID and existing equivalent tasks. A new state supplies a fresh loop ID. Never combine contradictory loops.

Create and verify

Create initial work with a shared goal and loop ID:

uuid="$("$CREATE_SKILL/scripts/dct-create.sh" --goal "$goal" --loop-id "$loop_id" \
  --description 'implement X' --acceptance 'X passes focused checks')"

Create findings with --from-task <producer>; add dependency, input branch, review ancestry, and an explicit round only when the default next round is not correct. Use --json when wiring a predicted review branch.

After final sync, export returned UUIDs and verify pending status, no start or assignee, identity, acceptance, routing, dependencies, and ancestry. Report the goal/loop identity and created UUIDs, stating that none were claimed.

Invoke dev-ask only when an environmental or harness failure occurs.

Finish

Every run ends with this command, including runs that created no task:

"$LOOP_SKILL/scripts/dl-finish.sh" tasks-created "$loop_id"

The single exception is a composed run: skip this only when dev-loop loaded this skill as a stage in the current session and will finish on your behalf. If you are not certain you are that case, you are not that case — run it.

Report first, then run it. Nothing follows it: no summary, no verification, no closing message. Reaching the end of your turn without it is an incomplete run, not a finished one — the worker process stays alive holding its queue, and the poll loop launches nothing until someone kills it by hand.

Preserve inherited AGENT_PID and AGENT_NOTIFY verbatim so the command can notify the controller and terminate the worker; never alter either value to make the helper return successfully.