| name | add-away |
| description | [ADD v0.11.0] Declare absence — get autonomous work plan for the duration |
| argument-hint | [duration, e.g. '4 hours', '30 minutes', 'end of day'] |
Codex interaction mode notice (ADD)
This skill depends on structured question/answer turns. Behavior depends on
Codex's current mode:
- Plan mode: call the
ask_user_question tool for each prompt below.
One question per call. Wait for the user's answer before moving on.
- Default mode (no
ask_user_question available): emit the questions
inline as a numbered list, then halt and wait for the user's next
prompt. Do not improvise, infer, or fabricate answers — this skill
fails closed if required input is missing. Resume only after the user
replies.
The skill body below defines what to ask; the shim only governs how to ask.
ADD Away Command v0.11.0
The human is stepping away. Establish what work can proceed autonomously and what must wait.
Phase 1: Understand the Absence
Parse the duration from $ARGUMENTS. If not provided, default to 2 hours. Do not ask — just acknowledge the default:
"No duration specified — I'll plan for a 2-hour session. Say /add-away 4 hours next time to adjust."
Phase 2: Assess Available Work
- Read
.add/config.json for autonomy level and environment tier
- Re-read
docs/prd.md to ground yourself in the project's objectives and scope — this keeps autonomous work aligned with the product vision
- Scan
specs/ for specs with status "Approved" or "Implementing"
- Scan
docs/plans/ for plans with status "Approved" or "In Progress"
- Check current git status for in-progress work
- Run TodoWrite to see current task list
Categorize Work
Autonomous (can do without human):
- Tasks with clear specs and plans where requirements are unambiguous
- Writing tests for specced features (RED phase)
- Implementing against existing failing tests (GREEN phase)
- Refactoring with test coverage (REFACTOR phase)
- Running quality gates and fixing lint/type errors
- Writing documentation for completed features
- Code review of existing PRs
Queued (needs human decision):
- Any task where the spec is ambiguous or missing
- Architecture decisions with multiple valid approaches
- Deployment to staging or production
- New feature specs (need interview)
- Dependency upgrades with breaking changes
- Anything that would benefit from a Decision Point
Phase 3: Present the Plan
Got it — you'll be away for approximately {DURATION}.
AUTONOMOUS WORK PLAN:
━━━━━━━━━━━━━━━━━━━
{numbered list of tasks, with spec references}
Estimated completion: {rough estimate}
QUEUED FOR YOUR RETURN:
━━━━━━━━━━━━━━━━━━━━━
{numbered list of decisions/tasks that need human input}
I'll maintain a work log and have a return briefing ready.
Phase 4: Get Confirmation
Ask: "Does this plan look right? Anything you want me to prioritize or avoid while you're away?"
Wait for confirmation before starting autonomous work.
Phase 5: Create Away Log
Write .add/away-log.md to track progress:
# Away Mode Log
**Started:** {timestamp}
**Expected Return:** {timestamp}
**Duration:** {duration}
## Work Plan
{the agreed plan}
## Progress Log
| Time | Task | Status | Notes |
|------|------|--------|-------|
Update this log as work progresses.
During Away Mode
Away mode grants elevated autonomy. The human is not available — do not wait for input on routine development tasks.
Autonomous Operations (do NOT ask for permission)
- Commit to feature branches — follow conventional commit format, commit after each TDD phase
- Push to feature branches — push regularly so work is not lost
- Create PRs — open PRs when a feature branch is ready for review
- Run and fix quality gates — lint, types, formatting errors are fixed without asking
- Read specs, plans, and PRD — re-read
docs/prd.md whenever you need to validate a decision against the product vision
- Run tests — execute test suites freely to verify your work
- Install dev dependencies — if tests or builds need a missing dev dependency, install it
- Promote through environments — follow the promotion ladder defined in the environment-awareness rule. If verification passes at one level and
autoPromote is true for the next environment, deploy there and verify. Rollback automatically on failure (see environment-awareness rule for details).
Boundaries (do NOT cross without human)
- Do NOT deploy to production or any environment where
autoPromote: false
- Do NOT merge PRs to main/production branches
- Do NOT start new features that lack a spec — if you finish all planned work, write documentation, improve test coverage, or refactor within existing specs
- Do NOT make irreversible changes (drop tables, delete branches, force push)
- Do NOT make architecture decisions with multiple valid approaches — log the decision point and move on
Staying Aligned
- Before starting each task, re-read the relevant spec and plan
- If a task feels ambiguous, check the PRD (
docs/prd.md) for guidance
- If still ambiguous after reading the PRD, log the question and skip to the next task — do not guess on product direction
Work Discipline
- ONLY work on tasks from the approved plan
- Log every completed task in
.add/away-log.md
- If a blocker is hit, log it and move to the next task
- Send brief status pulses in the conversation at reasonable intervals