Skip to content

gpsaggese/helpers / .claude/skills/auto_task.execute_interactively

auto_task.execute_interactively

Verified

Execute a sequence of tasks in a loop "execute -> review -> commit"

Goal

  • Execute an ordered list of well-specified, sequential tasks <FILE> by alternating agent execution, human review, and commit in a branch
  • Create the GH issue, branch, and PR for <FILE> before the loop: each auto_task maps to one GH issue and one branch/PR (see .claude/skills/auto_task.rules.md)
  • Follow .claude/skills/auto_task.rules.md for how tasks are queued, specified, and named

When to Use This Skill

  • Follow .claude/skills/auto_task.rules.md section "Ask for Clarification Before Executing an Unclear Plan" when a task's spec is unclear or incomplete

Conventions

  • Follow .claude/skills/auto_task.rules.md for queue, spec, naming, and coding/testing conventions

Constraints

  • Create the GH issue, branch, and (draft) PR for <FILE> up front: this is this skill's job, not the user's
  • Do not commit any code change for a task: committing is the user's decision
  • Do not merge any PR: merging is the user's decision

Workflow

Confirm the Task List

  • Read <FILE> and follow .claude/skills/auto_task.rules.md section "Confirm the Task List Before Executing" for the format and the problem/solution check
  • Make sure each task is clear. Add a plan of 5 markdown bullets under each task
  • If the order or a dependency is unclear, follow the "Ask for Clarification Before Executing an Unclear Plan" section of .claude/skills/auto_task.rules.md before starting

Create the Issue and the Branch

  • Follow .claude/skills/auto_task.rules.md section "One GitHub Issue Is the Unit of Work": one GH issue covers the whole task list in <FILE>

  • Create the issue and branch (<RepoPrefix>Task<IssueNum>_<Description>) in one shot with git_create_issue_and_branch.py. Unlike invoke gh_issue_create and invoke git_branch_create, it also branches submodules (e.g., helpers_root)

    > git_create_issue_and_branch.py \
        --gh_issue_title "<TITLE>" --gh_issue_body "<Description of the task list>" \
        --no_create_pr
    
    • Pass --gh_issue_id <NUM> instead of --gh_issue_title to reuse an issue
    • --no_create_pr: a PR needs a commit ahead of master. Open it after the first task's commit (see "Wait for User to Confirm")
  • git add (do not commit) <FILE> on the new branch, so its plan is tracked from the start, per @.claude/task_instructions.md

Loop Over the Tasks

Execute Each Task

  • Execute each task following @.claude/task_instructions.md
  • Keep <FILE> updated per .claude/skills/auto_task.rules.md section "Track Task Status"

Verify Each Task

  • If the N-th task writes code, test it locally and in CI following:
    • .claude/skills/github.get_pr_to_pass_ci_tests/SKILL.md
    • .claude/skills/github.get_pr_to_pass_local_tests/SKILL.md

Wait for User to Confirm

  • After each task, wait for the user to confirm, validate, and review, then commit

  • Push the commit. If the branch has no PR yet (first task), open the draft PR now, since there is now a commit to diff against master

    • If the task spans several repos (see .claude/skills/auto_task.rules.md "Multi-Repo Issues, Branches, and PRs"), open a draft PR in each affected repo, then refresh the issue's companion PR links:

      > git_create_issue_and_branch.py --gh_issue_id <NUM> --submodules \
          --update_pr_links
      
  • After the user confirms the N-th task, do the N+1 task the same way

Verification

  • One GH issue and one branch/PR were created for <FILE> before the loop, named per .claude/skills/auto_task.rules.md
  • Each task's status in <FILE> reflects its progress ([ ], [.], or [x])
  • Tests touched by each task pass locally and in CI before the task is marked done
  • No task started before its spec was clarified with the user
  • No commit was made without the user's confirmation
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 318kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 318kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 318kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 318kUpdated Claude CodeCodexVerifiedInstall