Wira Delta Indonesia

Daily Operations

Autonomous Daily Operations

Once the architecture is in place, everyday work runs as a daily rhythm through four skills you type inside your agent. This is the daily tier of WDI Method.

The Daily Tier Sequence

Daily work follows four steps:

[Hand-Test Notes / Bug Reports]
            │
            ▼  1. /wdi-daily-what-to-build
[Spec and tickets on the development branch, reviewed]
            │
            ▼  2. /wdi-daily-autopilot
[Mandate accepted (preflight if there is none),
 loop on branch autopilot/<mandate-id>,
 one PR ready for review]
            │
            │  The owner merges.
            ▼  3. /wdi-daily-what-to-test
[Hand-test checklist from the closed tickets]
            │
            ▼  4. /wdi-prune-or-archive
[Closed specs archived or pruned]

Skill 1: /wdi-daily-what-to-build

Turns hand-testing notes, QA observations, or bug reports into a reviewed spec or ticket on the development branch, for a later autopilot run:

/wdi-daily-what-to-build [reviewer] [--no-review] <notes>
  • Keeps your notes as written, then checks them against what is already promised and built.
  • Writes the spec or ticket through wdi-build's engines, to-spec and to-tickets.
  • Sends the draft to a read-only reviewer for a second opinion before code is written; --no-review skips it.
  • Stops. It never commits, pushes, or starts the autopilot.

Skill 2: /wdi-daily-autopilot

Checks for an accepted mandate, resolves reviewers from local config, and starts the loop:

/wdi-daily-autopilot [self-review] [peer] [interval] [--skip-peer-review | --no-review]
  • Runs the preflight if there is no accepted mandate yet; the loop starts only under one you accepted.
  • Starts the loop on the host's own scheduler, every 10 minutes by default: /loop 10m /wdi-autopilot on Claude Code, Command Code, and Qoder, and the host's scheduler on CodeBuddy, Cline, Goose, GitHub Copilot, Warp, and AdaL. On every other host it runs one iteration each time you type it, under the same mandate, ledger, and branch. It never falls back to a shell loop or an OS scheduler.
  • The loop runs on its own branch, autopilot/<mandate-id>, in an isolated worktree or in a clean checkout used only by that run. It never runs on a shared or dirty checkout.
  • The coordinating session writes the code itself, test-first, and records every decision in its ledger, .control/memlog/autopilot-<mandate-id>.md.
  • The run ends with one PR ready for review. The owner merges.

Skill 3: /wdi-daily-what-to-test

After a merge, prepares the next round of hand-testing:

/wdi-daily-what-to-test [web <target> | mobile <target> | desktop]
  • Syncs the development branch with a fast-forward pull.
  • Deletes local task branches already merged into the development branch, and their worktrees when clean. It deletes a remote branch only when the merge is confirmed, and never deletes the primary or development branch.
  • With the desktop target, it checks whether the app binary is still running before it rebuilds. It closes the app only if its own previous smoke run started it; otherwise it reports the PID and stops, so you can close it yourself. It never force-kills a process.
  • Builds a checklist from the tickets closed since the last sync (before_sync..HEAD). With no argument it only syncs, prunes, and builds the checklist.

Skill 4: /wdi-prune-or-archive

Retires closed specs through lifecycle.py, which checks first and rolls back on failure:

/wdi-prune-or-archive [--spec <id> | --all-closed] [--archive | --prune] [--dry-run]
  • --archive: moves closed specs from .scratch/ into .archive/specs/.
  • --prune: removes them with git rm.
  • Either way, the spec row stays in specs.yaml. Only specs with status closed qualify. With no argument, the skill asks which specs to handle.