pi-goal-list-loop-audit 0.15.1 → 0.17.0

package/README.md

@@ -33,13 +33,14 @@ Four top-level commands, that's all:
 /goal                              # drafting: agent grills, you Confirm
 /goal "audit the repo"             # no contract clause → agent grills you first (propose is gated on it)
 /goal "Step 1. Step 2. Done when: tests pass."   # has contract → starts now
+/goal start "fix the flaky login test"          # explicit skip-draft: starts now, no interview (auditor infers the contract)
 /goal status                       # show state
 /goal pause                        # pause
 /goal resume                       # resume
 /goal cancel                       # abort
 /goal tweak "<new objective>"      # edit in place (Confirm dialog)
 /goal archive                      # archived goals, newest first
-/gla                               # open the settings UI (or /gla key=value)
+/glla                               # open the settings UI (or /glla key=value)
 /list add                          # draft a contract (or a whole batch via items[])
 /list add "<objective>"            # queue one directly
 /list add plan.md                  # file detected → bulk import, one Confirm
@@ -69,7 +70,9 @@ your objectives can start with any verb.)
 
 Drafting rules: **no-args drafts, args-without-a-`Done when:`-clause get
 grilled by the agent (proposing is mechanically blocked until you have
-replied at least once), args-with-a-clause start instantly, a file path is
+replied at least once — typed chat or an answered `ask_user_question` dialog
+both count), args-with-a-clause start instantly, `/goal start` skips the
+interview by explicit command, a file path is
 bulk direct.** A
 sisyphus-style plan file (checklists, bullets, numbered, plain lines) imports
 as-is — headings become nothing, items become goals. And the drafter itself
@@ -148,7 +151,7 @@ Each loop is a different policy class on the same status machine.
 
 ## Live TUI (always know it's on)
 
-A persistent `gla:` status segment + an above-editor widget show the current
+A persistent `glla:` status segment + an above-editor widget show the current
 goal/loop at all times: objective, status, elapsed, tokens, next task or loop
 metric, pause reason, and live auditor progress during audits. If something is
 running, you can see it — no command needed.
@@ -163,19 +166,19 @@ No external watchdog plugin needed.
 ## Config (one global place, rarely opened)
 
 ```
-/gla                                # open the settings UI
-/gla model=provider/id              # auditor model override → GLOBAL
-/gla thinking=high                  # auditor thinking → GLOBAL
-/gla notify='cmd "$1"'              # push on complete/pause/stop → GLOBAL
-/gla tokenlimit=10000000            # per-goal token budget (default: off) → GLOBAL
-/gla tokenlimit=0                   # explicitly no cap (the default)
-/gla project tokenlimit=500         # rare per-project override
+/glla                                # open the settings UI
+/glla model=provider/id              # auditor model override → GLOBAL
+/glla thinking=high                  # auditor thinking → GLOBAL
+/glla notify='cmd "$1"'              # push on complete/pause/stop → GLOBAL
+/glla tokenlimit=10000000            # per-goal token budget (default: off) → GLOBAL
+/glla tokenlimit=0                   # explicitly no cap (the default)
+/glla project tokenlimit=500         # rare per-project override
 ```
 
 Resolution per key: **project > global > defaults**. The auditor defaults to
 your pi session model. When the session provider is extension-registered the
 auditor can't auth it — you're told once (info level) with the fix:
-`/gla model=provider/id`, set once, rarely touched again. The plugin never
+`/glla model=provider/id`, set once, rarely touched again. The plugin never
 picks a model itself. Thinking follows the session too (floor `high`).
 
 ## Token guard
@@ -183,7 +186,7 @@ picks a model itself. Thinking follows the session too (floor `high`).
 Every goal tracks real token usage; crossing the budget pauses the goal.
 Default 10,000,000 per goal — a runaway threshold, not a big-goal threshold
 (real research/feature goals legitimately burn 2-4M). Tune with
-`/gla tokenlimit=<n>`. Loop 3 doesn't need this cap — it has its own brakes
+`/glla tokenlimit=<n>`. Loop 3 doesn't need this cap — it has its own brakes
 (max iterations + plateau).
 
 ## Compatibility (what goes well, what conflicts)
@@ -209,8 +212,8 @@ not requirements).
 
 **Two footnotes**: (1) extension-registered providers work in the main session
 but not the auditor's extension-less session — if audits fail auth, set the
-override once with `/gla model=`. (2) `pi-notify-agent` notifies on every turn;
-`/gla notify=` fires only on goal complete/pause/loop stop.
+override once with `/glla model=`. (2) `pi-notify-agent` notifies on every turn;
+`/glla notify=` fires only on goal complete/pause/loop stop.
 
 ## Files
 

package/docs/DESIGN.md

@@ -179,7 +179,7 @@ active → aborted           (user /pi-gla-cancel)
 
 ### Decision 9: JSONL state (deterministic compaction)
 
-Goal state lives in `.pi-gla/active.jsonl`. Each line is a state transition. On compaction, the summary is rebuilt deterministically from the JSONL (autoresearch pattern).
+Goal state lives in `.pi-glla/active.jsonl`. Each line is a state transition. On compaction, the summary is rebuilt deterministically from the JSONL (autoresearch pattern).
 
 This protects against model-generated summaries losing fidelity.
 

package/extensions/goal-loop-auditor.ts

@@ -297,7 +297,7 @@ export async function runGoalCompletionAuditor(args: {
         output,
         model: modelLabel(model),
         thinkingLevel,
-        error: `Auditor produced no output${streamError ? `: ${streamError}` : " — the auditor session likely failed (check the model's auth/quota, or set a working one with /gla model=provider/id)"}`,
+        error: `Auditor produced no output${streamError ? `: ${streamError}` : " — the auditor session likely failed (check the model's auth/quota, or set a working one with /glla model=provider/id)"}`,
       };
     }
 

package/extensions/goal-loop-core.ts

@@ -158,10 +158,10 @@ export interface Goal {
 export type GoalRoute =
   | { kind: "draft" }
   | { kind: "set"; text: string }
-  | { kind: "sub"; name: "status" | "pause" | "resume" | "cancel" | "tweak" | "archive"; rest: string };
+  | { kind: "sub"; name: "status" | "pause" | "resume" | "cancel" | "tweak" | "archive" | "start"; rest: string };
 
 const GOAL_EXACT_SUBS = new Set(["status", "pause", "resume", "cancel"]);
-const GOAL_ARG_SUBS = new Set(["tweak", "archive"]);
+const GOAL_ARG_SUBS = new Set(["tweak", "archive", "start"]);
 
 export function routeGoalArgs(raw: string): GoalRoute {
   const trimmed = raw.trim();
@@ -173,7 +173,7 @@ export function routeGoalArgs(raw: string): GoalRoute {
     return { kind: "sub", name: first as "status" | "pause" | "resume" | "cancel", rest: "" };
   }
   if (GOAL_ARG_SUBS.has(first)) {
-    return { kind: "sub", name: first as "tweak" | "archive", rest };
+    return { kind: "sub", name: first as "tweak" | "archive" | "start", rest };
   }
   return { kind: "set", text: trimmed };
 }
@@ -334,7 +334,17 @@ export const DEFAULT_STATE: State = {
 // =================================================================
 
 export function piGlaDir(cwd: string): string {
-  return path.join(cwd, ".pi-gla");
+  const dir = path.join(cwd, ".pi-glla");
+  // v0.17.0: one-time migration of the pre-rename state dir (.pi-gla →
+  // .pi-glla). Active goals, ledgers, and project settings move with the
+  // name — no relics, no lost state.
+  const legacy = path.join(cwd, ".pi-gla");
+  try {
+    if (!fs.existsSync(dir) && fs.existsSync(legacy)) fs.renameSync(legacy, dir);
+  } catch {
+    // read-only fs or partial state — fall through and use the new dir
+  }
+  return dir;
 }
 
 export function goalMdPath(cwd: string, id: string): string {

package/extensions/goal-loop-display.ts

@@ -46,28 +46,28 @@ export interface AuditDisplayProgress {
 }
 
 /**
- * One-line status for ctx.ui.setStatus("pi-gla", …).
+ * One-line status for ctx.ui.setStatus("pi-glla", …).
  * Returns undefined when nothing is being supervised (clears the segment).
  */
 export function buildStatusText(state: State, audit?: AuditDisplayProgress | null, now = Date.now()): string | undefined {
   if (state.loop?.active) {
     const l = state.loop;
     const arrow = l.direction === "min" ? "↓" : "↑";
-    return `gla: loop ${arrow} iter ${l.iteration}/${l.maxIterations} · best ${l.bestValue ?? "n/a"} · stall ${l.stallCount}/${l.plateauWindow}`;
+    return `glla: loop ${arrow} iter ${l.iteration}/${l.maxIterations} · best ${l.bestValue ?? "n/a"} · stall ${l.stallCount}/${l.plateauWindow}`;
   }
   const g = state.goal;
   if (!g) return undefined;
   if (g.status === "auditing") {
     const tool = audit?.currentTool ? ` · ${audit.currentTool}` : "";
-    return `gla: auditing…${tool}`;
+    return `glla: auditing…${tool}`;
   }
   if (g.status === "paused") {
-    return `gla: paused ⏸ ${truncate(g.pauseReason ?? "", 40)}`;
+    return `glla: paused ⏸ ${truncate(g.pauseReason ?? "", 40)}`;
   }
   if (g.status === "active") {
     const queue = state.list?.length ? ` · list ${state.list.length}` : "";
     const tasks = g.taskList ? ` ${countDone(g)}/${countTotal(g)} tasks ·` : "";
-    return `gla: goal ●${tasks} ${fmtElapsed(now - Date.parse(g.createdAt))}${queue}`;
+    return `glla: goal ●${tasks} ${fmtElapsed(now - Date.parse(g.createdAt))}${queue}`;
   }
   return undefined; // complete/aborted → clear
 }
@@ -99,7 +99,7 @@ function countTotal(g: Goal): number {
 // ---- above-editor widget (multi-line panel) ----
 
 /**
- * Widget lines for ctx.ui.setWidget("pi-gla", lines).
+ * Widget lines for ctx.ui.setWidget("pi-glla", lines).
  * Returns undefined when nothing is worth showing.
  */
 export function buildWidgetLines(state: State, audit?: AuditDisplayProgress | null, now = Date.now()): string[] | undefined {

package/extensions/loops/goal.ts

@@ -11,7 +11,7 @@
  *   /goal "<objective>" | /goal (draft) | /goal status|pause|resume|cancel|tweak <text>|archive
  *   /list add|show|next|remove|clear
  *   /loop (draft) | /loop start|status|stop
- *   /gla (settings UI) | /gla key=value | /gla project key=value
+ *   /glla (settings UI) | /glla key=value | /glla project key=value
  */
 
 import * as fs from "node:fs";
@@ -142,8 +142,8 @@ let uiTicker: NodeJS.Timeout | null = null;
 function refreshUI(ctx: ExtensionContext): void {
   if (!ctx.hasUI) return;
   try {
-    ctx.ui.setStatus("pi-gla", buildStatusText(state, latestAuditProgress));
-    ctx.ui.setWidget("pi-gla", buildWidgetLines(state, latestAuditProgress));
+    ctx.ui.setStatus("pi-glla", buildStatusText(state, latestAuditProgress));
+    ctx.ui.setWidget("pi-glla", buildWidgetLines(state, latestAuditProgress));
   } catch {
     // stale ctx — next event refreshes
   }
@@ -434,7 +434,7 @@ async function startDrafting(ctx: ExtensionContext, target: "goal" | "list" | "l
   const [file, label, tool] = prompts[target]!;
   ctx.ui.notify(
     seed
-      ? `${label}: the objective has no "Done when:" clause — the agent will grill you about it first (nothing activates until you confirm).`
+      ? `${label}: the objective has no "Done when:" clause — the agent will grill you about it first (nothing activates until you confirm). Skip the interview entirely: /goal start <objective>.`
       : `${label} started. The agent will grill until the contract is concrete, then ${tool} opens a Confirm dialog. No work begins before confirmation.`,
     "info",
   );
@@ -481,6 +481,16 @@ async function cmdGoal(args: string, ctx: ExtensionContext): Promise<void> {
     if (route.name === "cancel") return cmdCancel(ctx);
     if (route.name === "tweak") return cmdTweak(route.rest, ctx);
     if (route.name === "archive") return cmdGoals(ctx);
+    // v0.16.0: /goal start <objective> — explicit skip-draft. Activates
+    // immediately, no interview, no "Done when:" heuristic. Symmetric
+    // with /loop start. The auditor infers the contract from the objective.
+    if (route.name === "start") {
+      if (!route.rest) {
+        ctx.ui.notify("Usage: /goal start <objective> — activates immediately, skipping the drafting interview. (Without start, an objective needs a 'Done when:' clause or it gets drafted first.)", "warning");
+        return;
+      }
+      return cmdSet(route.rest, ctx, true);
+    }
   }
   return cmdSet(route.kind === "set" ? route.text : "", ctx);
 }
@@ -489,7 +499,7 @@ async function cmdGoal(args: string, ctx: ExtensionContext): Promise<void> {
 // /goal: bypass drafting, start now (the only entry in v0.1.0)
 // =================================================================
 
-async function cmdSet(args: string, ctx: ExtensionContext): Promise<void> {
+async function cmdSet(args: string, ctx: ExtensionContext, skipDraft = false): Promise<void> {
   let raw = args.trim();
   // Users naturally quote the objective ("/goal \"do X\""); strip one layer of
   // surrounding matching quotes so they don't leak into the goal text.
@@ -507,7 +517,8 @@ async function cmdSet(args: string, ctx: ExtensionContext): Promise<void> {
   // v0.11.0: a contract-less objective gets drafted, not activated raw —
   // the pi-goal-x lesson: arg + Enter is worse than a 5-minute draft.
   // Include an explicit "Done when: …" clause to activate instantly.
-  if (goalArgsNeedDrafting(raw)) {
+  // v0.16.0: /goal start bypasses this by explicit user command.
+  if (!skipDraft && goalArgsNeedDrafting(raw)) {
     await startDrafting(ctx, "goal", raw);
     return;
   }
@@ -534,7 +545,7 @@ async function cmdStatus(ctx: ExtensionContext): Promise<void> {
     `Objective: ${g.objective}`,
     `Auto-continue: ${g.autoContinue ? "on" : "off"}`,
     `Iteration: ${iterationCounter}`,
-    `Tokens: ${(g.usage?.tokensUsed ?? 0).toLocaleString()}${(g.usage?.tokensLimit ?? 0) > 0 ? ` / ${(g.usage!.tokensLimit).toLocaleString()}` : " (no cap — /gla tokenlimit=<n> to set)"}`,
+    `Tokens: ${(g.usage?.tokensUsed ?? 0).toLocaleString()}${(g.usage?.tokensLimit ?? 0) > 0 ? ` / ${(g.usage!.tokensLimit).toLocaleString()}` : " (no cap — /glla tokenlimit=<n> to set)"}`,
   ];
   if (g.auditHistory && g.auditHistory.length > 0) {
     lines.push(`Audits: ${g.auditHistory.length} (${g.auditHistory.filter((v) => v.approved).length} approved)`);
@@ -837,7 +848,7 @@ function notifyExternal(ctx: ExtensionContext, message: string): void {
 }
 
 // =================================================================
-// Loop 3: /loop — metric-driven forever loop
+// Loop 3: /loop — metric-driven process loop (never completes)
 //
 // The anti-doorknob law: the loop only believes a number. The orchestrator
 // runs the user's measure command (via pi.exec) after every agent turn;
@@ -1412,7 +1423,7 @@ function registerAgentTools(pi: any, ctx: ExtensionContext): void {
     parameters: Type.Object({
       objective: Type.String({ description: "The clarified, concrete objective (single item) or a summary when items[] is used" }),
       verificationContract: Type.Optional(Type.String({ description: "Checkable done-criteria (commands, file states, test outcomes)" })),
-      items: Type.Optional(Type.Array(Type.String(), { description: "LIST drafting only: many objectives at once (e.g. 'queue these 50 things'). Each becomes a queue item; per-item 'Done when:' clauses are honored." })),
+      items: Type.Optional(Type.Array(Type.String(), { description: "LIST drafting only: many objectives at once (e.g. 'queue these 50 things'). Each becomes a list item; per-item 'Done when:' clauses are honored." })),
     }),
     async execute(_id, params, _signal, _onUpdate, execCtx) {
       const p = params as { objective: string; verificationContract?: string; items?: string[] };
@@ -2079,7 +2090,7 @@ async function cmdSettings(args: string, ctx: ExtensionContext): Promise<void> {
 // We detect duplicates at session start and warn loudly once.
 // =================================================================
 
-const OUR_COMMANDS = ["goal", "gla", "list", "queue", "loop"];
+const OUR_COMMANDS = ["goal", "glla", "list", "loop"];
 let collisionWarned = false;
 
 // Providers verified to exist in a bare (extension-less) session. The auditor
@@ -2149,7 +2160,7 @@ export default function (pi: ExtensionAPI): void {
   //   /loop  — the metric loop (draft|start|status|stop)
   //   /gla   — the settings UI (+ scriptable key=value)
   pi.registerCommand("goal", {
-    description: "Set/draft a goal, or /goal status|pause|resume|cancel|tweak <text>|archive. Objectives without a 'Done when:' clause are grilled into a contract first (nothing activates until you confirm); include the clause to start instantly.",
+    description: "Set/draft a goal, or /goal status|pause|resume|cancel|tweak <text>|archive|start <objective>. Objectives without a 'Done when:' clause are grilled into a contract first; include the clause or use /goal start to skip the interview and activate instantly.",
     handler: (args: string, ctx: ExtensionContext) => { rememberCtx(ctx); return cmdGoal(args, ctx); },
   });
   const settingsHandler = (args: string, ctx: ExtensionContext) => { rememberCtx(ctx); return cmdSettings(args, ctx); };
@@ -2157,23 +2168,12 @@ export default function (pi: ExtensionAPI): void {
     description: "Open the settings UI for goals, loops, lists, and the auditor. Scriptable form: /glla key=value · /glla project key=value",
     handler: settingsHandler,
   });
-  // /gla kept as an alias (renamed package, v0.15.0) — muscle memory is real.
-  pi.registerCommand("gla", {
-    description: "Alias for /glla (package renamed to pi-goal-list-loop-audit in v0.15.0).",
-    handler: settingsHandler,
-  });
   pi.registerCommand("list", {
     description: "Loop 2: the list of audited goals — order is the default, not the law. /list add <obj or file or paste> | /list show | /list next [n] | /list remove <n> | /list clear",
     handler: (args: string, ctx: ExtensionContext) => { rememberCtx(ctx); return cmdList(args, ctx); },
   });
-  // Alias for one release (renamed 0.10.0 back to /list — order isn't
-  // mandatory with subagents; it's a shopping list, not a FIFO).
-  pi.registerCommand("queue", {
-    description: "Alias for /list (renamed in v0.10.0). Same subcommands.",
-    handler: (args: string, ctx: ExtensionContext) => { rememberCtx(ctx); return cmdList(args, ctx); },
-  });
   pi.registerCommand("loop", {
-    description: "Loop 3: metric-driven forever loop. /loop start \"<target>\" measure=\"<cmd>\" direction=min|max [done=<value>] [window=5] [max=50] | /loop status | /loop stop",
+    description: "Loop 3: metric-driven process — it never completes. /loop start \"<target>\" measure=\"<cmd>\" direction=min|max [window=5] [max=50] [time=<hours>] [tokens=<budget>] [branch=1] | /loop status | /loop stop. 'Improve until X' is a /goal, not a loop.",
     handler: (args: string, ctx: ExtensionContext) => { rememberCtx(ctx); return cmdLoop(args, ctx); },
   });
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-goal-list-loop-audit",
-  "version": "0.15.1",
+  "version": "0.17.0",
   "description": "Goal. Loop. Audit. Done. — a pi-coding-agent extension that supervises long-running work, with isolated auditor on each completion. Beat bamboozling by design: the auditor runs in a fresh session with no extensions, no skills, no editor — only the read tools needed to verify your goal.",
   "license": "MIT",
   "author": "dracon",

package/prompts/goal-loop-continuation.md

@@ -43,7 +43,7 @@ You have `read`, `write`, `edit`, `bash`, `grep`, `find`, `ls`, and the goal too
 
 If the objective decomposes into milestones and no task list exists yet, call `propose_task_list` early — the user confirms it, then you track progress with `complete_task` / `update_task_status` as you go (not in a batch at the end). Limits: 20 tasks, 5 subtasks per task.
 
-When the agent calls any of these, the orchestrator tracks the call and persists state to `.pi-gla/active.jsonl`.
+When the agent calls any of these, the orchestrator tracks the call and persists state to `.pi-glla/active.jsonl`.
 
 ## TASK WORKFLOW
 

package/scripts/smoke.sh

@@ -2,7 +2,7 @@
 # pi-goal-list-loop-audit — live integration smoke
 #
 # Drives a real pi session in tmux against a scratch dir and asserts on the
-# .pi-gla ledger. This is the M2 "integration harness": it exercises the full
+# .pi-glla ledger. This is the M2 "integration harness": it exercises the full
 # loop (goal → agent work → complete_goal → isolated auditor → archive) with
 # real models, which unit tests cannot do.
 #
@@ -43,7 +43,7 @@ wait_for() { # wait_for <pattern> <timeout-s>
   return 1
 }
 ledger_has() { # ledger_has <jq-ish python expr substring>
-  python3 - "$1" "$WORK/.pi-gla/active.jsonl" <<'EOF'
+  python3 - "$1" "$WORK/.pi-glla/active.jsonl" <<'EOF'
 import json, sys
 needle, path = sys.argv[1], sys.argv[2]
 try:
@@ -81,7 +81,7 @@ case "$SCENARIO" in
     sleep 2
     if [ -f "$WORK/smoke.txt" ]; then pass "smoke.txt created"; else fail "smoke.txt missing"; fi
     if ledger_has '"approved":true'; then pass "ledger records approval"; else fail "ledger missing approval"; fi
-    if ls "$WORK/.pi-gla/archive/"*.md >/dev/null 2>&1; then pass "goal archived"; else fail "archive empty"; fi
+    if ls "$WORK/.pi-glla/archive/"*.md >/dev/null 2>&1; then pass "goal archived"; else fail "archive empty"; fi
     if ledger_has '"regressionShieldPassed":true'; then pass "regression_shield recorded"; else fail "shield outcome missing"; fi
     ;;
 
@@ -93,11 +93,11 @@ case "$SCENARIO" in
     if wait_for "approved by auditor" 120; then pass "item 1 approved"; else fail "item 1 not approved"; fi
     # wait for second archive file
     for i in $(seq 1 120); do
-      n=$(ls "$WORK/.pi-gla/archive/"*.md 2>/dev/null | wc -l)
+      n=$(ls "$WORK/.pi-glla/archive/"*.md 2>/dev/null | wc -l)
       [ "$n" -ge 2 ] && break
       sleep 1
     done
-    n=$(ls "$WORK/.pi-gla/archive/"*.md 2>/dev/null | wc -l)
+    n=$(ls "$WORK/.pi-glla/archive/"*.md 2>/dev/null | wc -l)
     if [ "$n" -ge 2 ]; then pass "both items archived ($n)"; else fail "only $n archived"; fi
     if [ -f "$WORK/a.txt" ] && [ -f "$WORK/b.txt" ]; then pass "both files created"; else fail "files missing"; fi
     if ledger_has '"list":[]'; then pass "list drained"; else fail "list not empty"; fi
@@ -119,7 +119,7 @@ case "$SCENARIO" in
     echo 5 > "$WORK/num.txt"
     NOTIFY_LOG="$WORK/notify.log"
     # project scope — never write test config into the user's GLOBAL settings
-    send "/gla project notify='echo \$1 >> $NOTIFY_LOG'"
+    send "/glla project notify='echo \$1 >> $NOTIFY_LOG'"
     sleep 4
     send '/loop start "Reduce the number in num.txt toward zero, never below 0" measure="cat num.txt" direction=min window=3 max=12'
     say "waiting for plateau stop (up to 300s)"