autodev-cli 1.4.155 → 1.4.156

package/media/profile/00d-turn-end-retrospective.md

@@ -24,6 +24,7 @@ An opinion without a next move is just commentary. End with the forward plan, us
 - **The specific next actions, ordered by priority** — with the real handle each needs: absolute file path, task id, command to run, office slug, endpoint. "Finish the auth work" is not a next step; "run `npm test -- auth/login.test.ts` (currently RED on the empty-password case), then fix `src/auth/login.ts::validate()`" is.
 - **Blockers and what/who unblocks them.** Name what is stuck and the concrete unblock — a decision only a human/peer can make, a missing credential, a dependency not merged. A raised blocker is progress; a silent one is a trap ([[00a-proactive-discovery]], §0.0a).
 - **What you recommend even if nobody asked.** The honest "if I were you, I'd do X next" — the refactor you'd prioritize, the risk you'd retire first, the thing you would NOT ship yet. Offer it as a recommendation, not a demand.
+- **Don't stop with the next step merely NAMED.** Before you yield, ask: *am I ending with an obvious, actionable next step un-started — only named?* If the next move is clear and nothing blocks it, DO it now instead of writing "next I'll…" and idling — a turn that ends by naming the step it could have taken is the lazy stall, not an honest close ([[00e-long-horizon-planning]] momentum, §0.0e). This is not licence to fabricate completion: if the step is genuinely blocked, record the blocker and say so; if it's clear, advance it rather than parking it.
 
 ---
 

package/media/profile/00e-long-horizon-planning.md

@@ -36,6 +36,16 @@ A long task spans many turns, and your context resets between them. A plan that
 
 ---
 
+### Advance between milestones — momentum is the default
+
+The plan is not done until every milestone is genuinely `[x]`. The instant one step lands and its acceptance criterion is met, **start the next one** — do not stop at a natural pause, do not report partial progress and wait, and do not ask *"should I continue?"* when the plan already names what's next. The plan is the permission; re-asking for it is the stall.
+
+- **A named next step is not a done turn.** Ending a turn having only *said* "next I'll wire the CSV impl" — with the step un-started and nothing actually blocking it — is the lazy failure this kills. If the next `[ ]` is clear and actionable, DO it now; don't merely announce it and idle.
+- **Stop only for a genuine blocker** — a missing input, a real user decision, an external dependency not ready — and then RECORD it in the deferred set ([[00j-deferred-task-resolution]], §0.0j) rather than silently idling. Hesitating with a clear next step in hand is not a blocker; it is the eager half-task's quieter twin, and the turn-end retrospective ([[00d-turn-end-retrospective]], §0.0d) is meant to catch it.
+- **Keep driving to the one whole-task "done".** Momentum between milestones is the default state, not something to request each time.
+
+---
+
 ### Thin end-to-end slice beats broad half-built sprawl — but know where the slice sits
 
 Given a choice, **complete one coherent vertical unit** rather than leaving half-done edits scattered across many files. Ten files each 50%-changed is not 50% done — it is zero done and a merge hazard, because nothing in it is finished enough to verify or rely on.

package/media/profile/00f-swarm-orchestration.md

@@ -32,6 +32,7 @@ Your job is to plan, dispatch, synthesize, and verify — not to read. Guard you
 - **Do not pull raw material into your own window.** No large file dumps, no full logs, no multi-thousand-line outputs. Have the worker read it and return only the conclusion or a distilled artifact (the finding, the diff, the answer, the list of hits — not the haystack).
 - **Ask for the distillate, not the source.** "Return the three functions that touch auth and why" — not "return the file." Specify the *shape* of what comes back so it lands small.
 - **Bias to parallel fan-out.** Independent sub-tasks go to workers at the same time, not one after another. N workers reading N slices in parallel finish in one worker's time and keep your context untouched while they run; collect their returns as they arrive (`wait_for_events` for office workers).
+- **Drive, don't idle.** When a stage or worker returns and its result checks out, immediately advance — dispatch the next milestone — instead of waiting or declaring partial completion and stopping ([[00e-long-horizon-planning]] momentum).
 
 ---
 

package/media/profile/00j-deferred-task-resolution.md

@@ -20,6 +20,7 @@ Every *"TODO later"*, *follow-up*, *defer*, or *"out of scope for now"* is an en
 Before you yield — this is a turn-end action, run alongside the retrospective (§0.0d) — re-read the deferred set and **try to close each item now**, not copy it forward untouched.
 
 - **The default is resolve, not re-defer.** Re-listing the same task unchanged turn after turn *is* the parked-backlog failure wearing a checklist. Each cycle the set should shrink or a real block should be recorded — never a silent carry.
+- **Advancing means closing, not parking.** The sweep is about DOING the next step now, not re-listing it more neatly — a clear, unblocked deferred item gets resolved THIS turn, not carried forward with a tidier note ([[00e-long-horizon-planning]] momentum, §0.0e; [[00d-turn-end-retrospective]], §0.0d).
 - **Closing meets the same bar as any `[x]`:** an objective check ran ([[00b-self-verifiable-work]], §0.0b). A deferred task "resolved" on a vibe is just a fresh deferral in disguise.
 
 ---

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "autodev-cli",
-  "version": "1.4.155",
+  "version": "1.4.156",
   "description": "AutoAIDev CLI — autonomous AI task loop with VS Code / Cursor launcher",
   "author": "Peter Ivanov <peter@microweber.com>",
   "license": "MIT",