@colin4k1024/tsp 2.5.3 → 2.5.4

package/commands/goal.md

@@ -114,7 +114,13 @@ When ANY budget limit is hit, the goal escalates rather than continuing blindly.
 
 ## State Persistence
 
-Goals persist to `~/.claude/goals/{goalId}.json` (follows goal.schema.json).
+Goals persist through the shared loop state store:
+
+- `TSP_LOOP_STATE_DIR/goals/{goalId}.json` when `TSP_LOOP_STATE_DIR` is set
+- `.tsp/loops/goals/{goalId}.json` for project-local loop state
+- target defaults such as `~/.claude/loops/goals/{goalId}.json` or `~/.codex/loops/goals/{goalId}.json`
+
+Legacy `~/.claude/goals/{goalId}.json` remains readable during migration.
 
 - Active goals survive session restarts
 - SessionStart hook displays active goal reminder

package/commands/heartbeat.md

@@ -51,7 +51,39 @@ Execute all configured scans immediately (one-shot, no scheduling).
 
 ## Configuration
 
-Heartbeat reads from `.claude/heartbeat.yaml` in the project root:
+Heartbeat reads configuration in this order:
+
+1. `.tsp/loop.yaml` — preferred Loop Engineering spec. Gates become discovery scans.
+2. `.tsp/heartbeat.yaml` — heartbeat-only config.
+3. `.claude/heartbeat.yaml` — legacy compatibility path.
+
+Preferred `.tsp/loop.yaml`:
+
+```yaml
+loop:
+  id: ci-triage
+  description: Keep CI failures triaged and fix machine-checkable failures.
+  cadence: 30m
+  skill: loop-ci-triage
+  stateFile: .tsp/loops/state/ci-triage.md
+  gates:
+    - name: library-validation
+      command: node scripts/validate-library.js
+    - name: tests
+      command: npm test
+  maker:
+    role: backend-engineer
+    writeAccess: true
+  checker:
+    role: qa-engineer
+    writeAccess: false
+  budget:
+    maxIterations: 10
+    maxDuration: 2h
+    maxDollars: 5
+```
+
+Heartbeat-only config remains supported:
 
 ```yaml
 heartbeat:
@@ -119,6 +151,7 @@ Each scan result is classified into an action:
 - **`/triage`**: Heartbeat populates triage inbox from `triage` scan failures
 - **Budget**: Integrates with cost tracking; pauses if hourly budget exceeded
 - **Hooks**: Uses CronCreate/CronDelete for scheduling (7-day auto-expiry applies)
+- **State**: Writes via the shared loop state store under `TSP_LOOP_STATE_DIR`, `.tsp/loops/`, or a target default
 
 ## Arguments
 

package/commands/loop-start.md

@@ -4,29 +4,57 @@ Start a managed autonomous loop pattern with safety defaults.
 
 ## Usage
 
-`/loop-start [pattern] [--mode safe|fast]`
+`/loop-start [pattern] [--mode safe|fast] [--spec .tsp/loop.yaml]`
 
 - `pattern`: `sequential`, `continuous-pr`, `rfc-dag`, `infinite`
 - `--mode`:
   - `safe` (default): strict quality gates and checkpoints
   - `fast`: reduced gates for speed
+- `--spec`: loop spec path, default `.tsp/loop.yaml`
+
+## Loop Engineering Intake
+
+Do not start a scheduled loop until all four conditions hold:
+
+1. The task repeats often enough to amortize loop setup.
+2. Verification is automated with at least one hard gate.
+3. Budget is explicit: iterations, wall-clock time, and cost.
+4. Tool access is sufficient and bounded for the intended work.
+
+If any condition fails, use `/quick`, `/verify`, or a one-shot script instead of
+creating a loop.
 
 ## Flow
 
 1. Confirm repository state and branch strategy.
-2. Select loop pattern and model tier strategy.
-3. Enable required hooks/profile for the chosen mode.
-4. Create loop plan and write runbook under `.claude/plans/`.
-5. Print commands to start and monitor the loop.
+2. Validate `.tsp/loop.yaml` against `schemas/loop-spec.schema.json`.
+3. Initialize loop state under `TSP_LOOP_STATE_DIR`, `.tsp/loops/`, or the target adapter default.
+4. Select loop pattern and model tier strategy.
+5. Enable required hooks/profile for the chosen mode.
+6. Print commands to start and monitor the loop.
 
 ## Required Safety Checks
 
 - Verify tests pass before first loop iteration.
 - Ensure `ECC_HOOK_PROFILE` is not disabled globally.
-- Ensure loop has explicit stop condition.
+- Ensure loop has at least one objective gate.
+- Ensure loop has explicit stop conditions and budget limits.
+- Ensure maker and checker are separate roles or agents.
+- Ensure loop writes go through a branch or worktree, not protected main.
+
+## State
+
+Loop runtime state is target-neutral:
+
+- `TSP_LOOP_STATE_DIR` when set
+- project-local `.tsp/loops/`
+- target default, such as `~/.claude/loops/` or `~/.codex/loops/`
+
+Legacy Claude goal and triage paths remain readable for migration.
 
 ## Arguments
 
 $ARGUMENTS:
 - `<pattern>` optional (`sequential|continuous-pr|rfc-dag|infinite`)
 - `--mode safe|fast` optional
+- `--spec <path>` optional

package/commands/triage.md

@@ -86,7 +86,14 @@ Items enter triage from:
 
 ## Storage
 
-Inbox is stored as newline-delimited JSON at `~/.claude/triage/inbox.jsonl`.
+Inbox is stored as newline-delimited JSON through the shared loop state store:
+
+1. `TSP_LOOP_STATE_DIR/triage/inbox.jsonl` when `TSP_LOOP_STATE_DIR` is set
+2. `.tsp/loops/triage/inbox.jsonl` for project-local state
+3. target defaults such as `~/.claude/loops/triage/inbox.jsonl` or `~/.codex/loops/triage/inbox.jsonl`
+
+Legacy `~/.claude/triage/inbox.jsonl` remains a migration source for older Claude installs.
+
 This format is:
 - Append-only (safe for concurrent writes)
 - Human-readable (one JSON object per line)

package/hooks/README.md

@@ -26,7 +26,7 @@ User request → Claude picks a tool → PreToolUse hook runs → Tool executes
 | **Git push reminder** | `Bash` | Reminds to review changes before `git push` | 0 (warns) |
 | **Pre-commit quality check** | `Bash` | Runs quality checks before `git commit`: lints staged files, validates commit message format when provided via `-m/--message`, detects console.log/debugger/secrets | 2 (blocks critical) / 0 (warns) |
 | **Doc file warning** | `Write` | Warns about non-standard `.md`/`.txt` files (allows README, CLAUDE, CONTRIBUTING, CHANGELOG, LICENSE, SKILL, docs/, skills/); cross-platform path handling | 0 (warns) |
-| **Strategic compact** | `*` | Suggests `/compact` when context usage crosses 70/85/95% thresholds, using Claude context-window metrics when available | 0 (warns) |
+| **Strategic compact** | `*` | Suggests `/compact` when context usage crosses 65/70/85/95% thresholds, using CCometixLine-compatible remaining context first | 0 (warns) |
 | **InsAIts security monitor (opt-in)** | `Bash\|Write\|Edit\|MultiEdit` | Optional security scan for high-signal tool inputs. Disabled unless `ECC_ENABLE_INSAITS=1`. Blocks on critical findings, warns on non-critical, and writes audit log to `.insaits_audit_session.jsonl`. JS wrapper is the canonical hook entry; the Python monitor is an explicit third-party SDK exception. Requires `pip install insa-its`. [Details](../scripts/hooks/insaits-security-monitor.py) | 2 (blocks critical) / 0 (warns) |
 
 ### PostToolUse Hooks
@@ -48,7 +48,7 @@ User request → Claude picks a tool → PreToolUse hook runs → Tool executes
 |------|-------|-------------|
 | **Session start** | `SessionStart` | Loads previous context and detects package manager |
 | **PUA always-on restore** | `SessionStart` | Restores flavor, level, and failure count when `~/.claude/pua/config.json` enables always-on |
-| **Pre-compact** | `PreCompact` | Saves state before context compaction |
+| **Pre-compact** | `PreCompact` | Saves state and increments project/session compact count before context compaction |
 | **PUA state snapshot** | `PreCompact` | Writes flavor, level, and failure count to the PUA builder journal |
 | **Console.log audit** | `Stop` | Checks all modified files for `console.log` after each response |
 | **Session summary** | `Stop` | Persists session state when transcript path is available |

package/hooks/harness-statusline.js

@@ -15,9 +15,7 @@
 const fs = require('fs');
 const path = require('path');
 const os = require('os');
-
-// Autocompact buffer ~16.5% — GSD pattern, normalize to usable context
-const AUTO_COMPACT_BUFFER_PCT = 16.5;
+const { resolveContextMetrics } = require('../scripts/lib/context-window');
 
 let input = '';
 // Guard: if stdin stalls (Windows/pipe issues), exit silently
@@ -31,8 +29,6 @@ process.stdin.on('end', () => {
     const session = data.session_id || '';
     const model = data.model?.display_name || 'Claude';
     const dir = data.cwd || data.workspace?.current_dir || process.cwd();
-    const remaining = data.context_window?.remaining_percentage;
-
     // --- Resolve active role from harness state file ---
     const claudeDir = process.env.CLAUDE_CONFIG_DIR || path.join(os.homedir(), '.claude');
     const projectDir = process.env.CLAUDE_PROJECT_DIR || dir;
@@ -49,30 +45,9 @@ process.stdin.on('end', () => {
     // --- Context window metrics ---
     let ctxBar = '';
     let usedPct = null;
-    if (remaining != null) {
-      const usableRemaining = Math.max(0, ((remaining - AUTO_COMPACT_BUFFER_PCT) / (100 - AUTO_COMPACT_BUFFER_PCT)) * 100);
-      usedPct = Math.max(0, Math.min(100, Math.round(100 - usableRemaining)));
-    } else if (data.transcript_path) {
-      // Parse actual token usage from transcript JSONL (CCometixLine approach)
-      try {
-        const { resolveTranscriptMetrics } = require('../scripts/lib/transcript-usage');
-        const modelId = (data.model && data.model.id) || process.env.CLAUDE_MODEL || null;
-        const metrics = resolveTranscriptMetrics(data.transcript_path, modelId);
-        if (metrics) {
-          usedPct = Math.max(0, Math.min(100, Math.round(metrics.usagePct)));
-        }
-      } catch (_) { /* ignore */ }
-
-      // Final fallback: file size estimate
-      if (usedPct == null) {
-        try {
-          const stat = fs.statSync(data.transcript_path);
-          const estimatedTokens = Math.round(stat.size * 0.25);
-          if (estimatedTokens > 0) {
-            usedPct = Math.max(0, Math.min(100, Math.round((estimatedTokens / 200000) * 100)));
-          }
-        } catch (_) { /* ignore */ }
-      }
+    const metrics = resolveContextMetrics(data);
+    if (metrics) {
+      usedPct = metrics.usagePct;
     }
 
     // Write bridge file for downstream hooks (suggest-compact, context-monitor)
@@ -81,8 +56,12 @@ process.stdin.on('end', () => {
         const bridgePath = path.join(os.tmpdir(), `harness-ctx-${session}.json`);
         fs.writeFileSync(bridgePath, JSON.stringify({
           session_id: session,
-          remaining_percentage: remaining != null ? remaining : null,
+          remaining_percentage: metrics?.remainingPct ?? null,
+          remaining_tokens: metrics?.remainingTokens ?? null,
           used_pct: usedPct,
+          context_limit: metrics?.contextLimit ?? null,
+          compact_count: metrics?.compactCount ?? null,
+          context_source: metrics?.source ?? null,
           active_role: activeRole,
           timestamp: Math.floor(Date.now() / 1000),
         }));

package/hooks/strategic-compact/README.md

@@ -13,12 +13,9 @@
 **触发时机**: 上下文压缩前自动调用
 
 **功能**:
-- 分析所有上下文元素
-- 按重要性分类：
-  - `keep`: 高价值内容（决策、结论、pending items、验证结果）
-  - `summarize`: 中等价值内容（长输出、对话历史）
-  - `discard`: 低价值内容（工具结果、搜索结果、重复信息）
-- 生成摘要指令
+- 保存 compaction 事件
+- 递增 `.tsp/context/compact-state.json` 中的 session compact count 与 total compact count
+- 把 compact 轮次写入 session tmp 文件，供后续恢复和动态上下文策略使用
 
 **优先级规则**:
 - 总是保留的类型：`decision`, `conclusion`, `pending_item`, `task_output`, `verification_result`, `error_fix`
@@ -27,13 +24,14 @@
 
 ### suggest-compact.js
 
-**触发时机**: `hooks/hooks.json` 的 `pre:all:strategic-compact` 在工具调用前运行；当真实上下文使用超过 70% 时注入 `/compact` 建议。
+**触发时机**: `hooks/hooks.json` 的 `pre:all:strategic-compact` 在工具调用前运行；当真实上下文使用超过 65% 时注入 `/compact` 建议。
 
 **功能**:
-- 优先读取 Claude hook 输入里的 `context_window.used_percentage` / `context_window.remaining_percentage`
-- 其次读取 `CLAUDE_CONTEXT_SIZE` / `CLAUDE_CONTEXT_LIMIT`
-- 最后读取 `harness-statusline.js` 写入的 `/tmp/harness-ctx-{session_id}.json`
-- 评估紧迫度：`low` (< 70%) | `medium` (70-85%) | `high` (85-95%) | `critical` (> 95%)
+- 优先读取 CCometixLine-compatible remaining context：`TSP_CONTEXT_WINDOW_JSON` / `CCOMETIXLINE_CONTEXT_JSON`、`TSP_CONTEXT_WINDOW_FILE` / `CCOMETIXLINE_CONTEXT_FILE`、hook 输入的 `ccometixline.context_window` / `ccometixline_context_window`
+- 其次读取 Claude hook 输入里的 `context_window.used_percentage` / `context_window.remaining_percentage`
+- 再退回 transcript JSONL usage、`CLAUDE_CONTEXT_SIZE` / `CLAUDE_CONTEXT_LIMIT`、`harness-statusline.js` 写入的 `/tmp/harness-ctx-{session_id}.json`
+- 评估紧迫度：`low` (< 65%) | `advisory` (65-70%) | `medium` (70-85%) | `high` (85-95%) | `critical` (> 95%)
+- 输出 `compact_count`，让动态上下文策略知道当前已经经历过几轮 compact
 - 估算可节省的 token
 - 提供具体压缩建议：
   - 压缩早期对话历史
@@ -59,7 +57,8 @@
 
 | 使用率 | 紧迫度 | 建议操作 |
 |--------|--------|---------|
-| < 70% | low | 无需操作 |
+| < 65% | low | 无需操作 |
+| 65-70% | advisory | 控制上下文增长 |
 | 70-85% | medium | 建议压缩，可选择性执行 |
 | 85-95% | high | 强烈建议压缩 |
 | > 95% | critical | 必须立即压缩 |

package/manifests/install-components.json

@@ -193,6 +193,14 @@
         "orchestration"
       ]
     },
+    {
+      "id": "capability:loop-engineering",
+      "family": "capability",
+      "description": "Loop Engineering runtime: automation, skills, target-neutral state, hard gates, and maker/checker loop primitives.",
+      "modules": [
+        "loop-engineering"
+      ]
+    },
     {
       "id": "capability:workflow-engine",
       "family": "capability",
@@ -396,6 +404,38 @@
         "workflow-quality"
       ]
     },
+    {
+      "id": "skill:loop-heartbeat",
+      "family": "skill",
+      "description": "Scheduled discovery automation that routes findings to goals or triage.",
+      "modules": [
+        "loop-engineering"
+      ]
+    },
+    {
+      "id": "skill:goal-convergence",
+      "family": "skill",
+      "description": "Goal convergence patterns for bounded autonomous loops.",
+      "modules": [
+        "loop-engineering"
+      ]
+    },
+    {
+      "id": "skill:rework-loop",
+      "family": "skill",
+      "description": "Rework loop patterns for checker-driven iteration.",
+      "modules": [
+        "loop-engineering"
+      ]
+    },
+    {
+      "id": "skill:autonomous-loops",
+      "family": "skill",
+      "description": "Autonomous loop design and safety guidance.",
+      "modules": [
+        "loop-engineering"
+      ]
+    },
     {
       "id": "skill:strategic-compact",
       "family": "skill",

package/manifests/install-modules.json

@@ -528,6 +528,49 @@
       "cost": "medium",
       "stability": "beta"
     },
+    {
+      "id": "loop-engineering",
+      "kind": "orchestration",
+      "description": "Loop Engineering command surface, target-neutral state runtime, loop spec schema, heartbeat discovery, and maker/checker loop skills.",
+      "paths": [
+        ".tsp/loop.example.yaml",
+        "commands/goal.md",
+        "commands/heartbeat.md",
+        "commands/triage.md",
+        "commands/loop-start.md",
+        "commands/loop-status.md",
+        "docs/runbooks/loop-engineering-usage.md",
+        "schemas/loop-spec.schema.json",
+        "scripts/lib/loop-state-store.js",
+        "scripts/lib/loop-spec.js",
+        "scripts/lib/loop-oracle.js",
+        "scripts/lib/completion-oracle.js",
+        "scripts/lib/heartbeat-scheduler.js",
+        "scripts/hooks/session-start-goal-resume.js",
+        "skills/loop-heartbeat",
+        "skills/goal-convergence",
+        "skills/rework-loop",
+        "skills/context-engineering",
+        "skills/context-lifecycle",
+        "templates/context-docs"
+      ],
+      "targets": [
+        "claude",
+        "cursor",
+        "codex",
+        "opencode",
+        "cangming",
+        "codebuddy",
+        "codewhale"
+      ],
+      "dependencies": [
+        "commands-core",
+        "platform-configs"
+      ],
+      "defaultInstall": false,
+      "cost": "medium",
+      "stability": "beta"
+    },
     {
       "id": "workflow-engine",
       "kind": "workflow",

package/manifests/install-profiles.json

@@ -78,6 +78,7 @@
         "workflow-engine",
         "workflow-defaults",
         "orchestration",
+        "loop-engineering",
         "swift-apple",
         "agentic-patterns",
         "devops-infra",
@@ -105,6 +106,7 @@
         "knowledge-graph",
         "workflow-engine",
         "workflow-defaults",
+        "loop-engineering",
         "design-prototyping",
         "enhanced-workflows",
         "rtk-optimization"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@colin4k1024/tsp",
-  "version": "2.5.3",
+  "version": "2.5.4",
   "description": "Open-source Team Skills Platform for role-based AI delivery workflows, shared skills, hooks, commands, and multi-platform installs.",
   "repository": {
     "type": "git",

package/schemas/loop-spec.schema.json

@@ -0,0 +1,124 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://colin4k1024.github.io/tsp/schemas/loop-spec.schema.json",
+  "title": "TSP Loop Engineering Spec",
+  "type": "object",
+  "required": ["loop"],
+  "additionalProperties": false,
+  "properties": {
+    "loop": {
+      "type": "object",
+      "required": [
+        "id",
+        "description",
+        "cadence",
+        "skill",
+        "stateFile",
+        "gates",
+        "maker",
+        "checker",
+        "budget"
+      ],
+      "additionalProperties": false,
+      "properties": {
+        "id": {
+          "type": "string",
+          "minLength": 1
+        },
+        "description": {
+          "type": "string",
+          "minLength": 1
+        },
+        "cadence": {
+          "type": "string",
+          "pattern": "^[0-9]+[mhd]$"
+        },
+        "skill": {
+          "type": "string",
+          "minLength": 1
+        },
+        "stateFile": {
+          "type": "string",
+          "minLength": 1
+        },
+        "gates": {
+          "type": "array",
+          "minItems": 1,
+          "items": {
+            "type": "object",
+            "required": ["name", "command"],
+            "additionalProperties": false,
+            "properties": {
+              "name": {
+                "type": "string",
+                "minLength": 1
+              },
+              "command": {
+                "type": "string",
+                "minLength": 1
+              },
+              "description": {
+                "type": "string"
+              }
+            }
+          }
+        },
+        "maker": {
+          "$ref": "#/$defs/actor"
+        },
+        "checker": {
+          "$ref": "#/$defs/actor"
+        },
+        "budget": {
+          "type": "object",
+          "required": ["maxIterations", "maxDuration", "maxDollars"],
+          "additionalProperties": false,
+          "properties": {
+            "maxIterations": {
+              "type": "number",
+              "exclusiveMinimum": 0
+            },
+            "maxDuration": {
+              "type": "string",
+              "pattern": "^[0-9]+[mh]$"
+            },
+            "maxDollars": {
+              "type": "number",
+              "exclusiveMinimum": 0
+            }
+          }
+        },
+        "escalation": {
+          "type": "object",
+          "additionalProperties": false,
+          "properties": {
+            "onBudgetExhausted": {
+              "type": "string",
+              "enum": ["triage", "human", "stop"]
+            },
+            "onSecurityFinding": {
+              "type": "string",
+              "enum": ["human", "triage", "stop"]
+            }
+          }
+        }
+      }
+    }
+  },
+  "$defs": {
+    "actor": {
+      "type": "object",
+      "required": ["role", "writeAccess"],
+      "additionalProperties": false,
+      "properties": {
+        "role": {
+          "type": "string",
+          "minLength": 1
+        },
+        "writeAccess": {
+          "type": "boolean"
+        }
+      }
+    }
+  }
+}

package/scripts/hooks/pre-compact.js

@@ -18,8 +18,16 @@ const {
   appendFile,
   log
 } = require('../lib/utils');
+const { recordCompactEvent } = require('../lib/context-window-state');
+
+async function main(rawInput = '{}') {
+  let input = {};
+  try {
+    input = JSON.parse(rawInput || '{}');
+  } catch (_) {
+    input = {};
+  }
 
-async function main() {
   const sessionsDir = getSessionsDir();
   const compactionLog = path.join(sessionsDir, 'compaction-log.txt');
 
@@ -27,7 +35,18 @@ async function main() {
 
   // Log compaction event with timestamp
   const timestamp = getDateTimeString();
-  appendFile(compactionLog, `[${timestamp}] Context compaction triggered\n`);
+  let compactEvent = null;
+  try {
+    compactEvent = recordCompactEvent(input, {
+      cwd: input.cwd || input.workspace?.current_dir || process.cwd(),
+    });
+  } catch (_) {
+    compactEvent = null;
+  }
+  const compactCountSuffix = compactEvent
+    ? ` (session compact #${compactEvent.sessionCompactCount}, total #${compactEvent.totalCompactCount})`
+    : '';
+  appendFile(compactionLog, `[${timestamp}] Context compaction triggered${compactCountSuffix}\n`);
 
   // If there's an active session file, note the compaction
   const sessions = findFiles(sessionsDir, '*-session.tmp');
@@ -35,14 +54,26 @@ async function main() {
   if (sessions.length > 0) {
     const activeSession = sessions[0].path;
     const timeStr = getTimeString();
-    appendFile(activeSession, `\n---\n**[Compaction occurred at ${timeStr}]** - Context was summarized\n`);
+    const countNote = compactEvent ? ` (compact #${compactEvent.sessionCompactCount})` : '';
+    appendFile(activeSession, `\n---\n**[Compaction occurred at ${timeStr}${countNote}]** - Context was summarized\n`);
   }
 
-  log('[PreCompact] State saved before compaction');
+  log(`[PreCompact] State saved before compaction${compactCountSuffix}`);
   process.exit(0);
 }
 
-main().catch(err => {
-  console.error('[PreCompact] Error:', err.message);
-  process.exit(0);
-});
+if (require.main === module) {
+  let input = '';
+  const stdinTimeout = setTimeout(() => main(input), 5000);
+  process.stdin.setEncoding('utf8');
+  process.stdin.on('data', chunk => { input += chunk; });
+  process.stdin.on('end', () => {
+    clearTimeout(stdinTimeout);
+    main(input).catch(err => {
+      console.error('[PreCompact] Error:', err.message);
+      process.exit(0);
+    });
+  });
+}
+
+module.exports = { main };

package/scripts/hooks/session-start-goal-resume.js

@@ -11,34 +11,17 @@
  * session start output so the user knows they can resume with `/goal resume`.
  *
  * Behavior:
- *   1. Scans ~/.claude/goals/ for goal files with state "active" or "paused"
+ *   1. Scans the TSP loop state goals store for active or paused goal files
  *   2. If found, appends a summary to the hook output
  *   3. Non-blocking: failures are silently ignored
  */
 
 const fs = require('fs');
-const path = require('path');
-
-function getGoalsDir() {
-  const home = process.env.HOME || process.env.USERPROFILE || '';
-  return path.join(home, '.claude', 'goals');
-}
+const { listGoals } = require('../lib/loop-state-store');
 
 function scanActiveGoals() {
-  const dir = getGoalsDir();
-  if (!fs.existsSync(dir)) return [];
-
   try {
-    return fs.readdirSync(dir)
-      .filter(f => f.endsWith('.json'))
-      .map(f => {
-        try {
-          return JSON.parse(fs.readFileSync(path.join(dir, f), 'utf-8'));
-        } catch {
-          return null;
-        }
-      })
-      .filter(g => g && (g.state === 'active' || g.state === 'paused'))
+    return [...listGoals('active'), ...listGoals('paused')]
       .sort((a, b) => new Date(b.updatedAt) - new Date(a.updatedAt));
   } catch {
     return [];