fix: token budget tracker for context overflow (#925)

fix: restore _format_error helper for test compatibility (#916)

agent/token_budget.py

@@ -0,0 +1,165 @@
+"""Token Budget — Poka-yoke guard against context overflow.
+
+Progressive warning system with circuit breakers:
+- 60%: Log warning, suggest summarization
+- 80%: Auto-compress, drop raw tool outputs
+- 90%: Block verbose tools, force wrap-up
+- 95%: Graceful termination with summary
+
+Usage:
+    from agent.token_budget import TokenBudget
+    budget = TokenBudget(max_tokens=128000)
+    budget.record_usage(prompt_tokens=500, completion_tokens=200)
+    status = budget.check()
+    # status.level: ok, warning, compress, block, terminate
+"""
+
+from __future__ import annotations
+
+import logging
+import time
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Any, Dict, List, Optional
+
+logger = logging.getLogger(__name__)
+
+
+class BudgetLevel(Enum):
+    """Token budget alert levels."""
+    OK = "ok"                 # < 60%
+    WARNING = "warning"       # 60-80%
+    COMPRESS = "compress"     # 80-90%
+    BLOCK = "block"           # 90-95%
+    TERMINATE = "terminate"   # > 95%
+
+
+@dataclass
+class BudgetStatus:
+    """Current budget status."""
+    level: BudgetLevel
+    used_tokens: int
+    max_tokens: int
+    percentage: float
+    remaining: int
+    message: str
+    actions: List[str] = field(default_factory=list)
+
+
+# Default thresholds
+THRESHOLDS = {
+    BudgetLevel.WARNING: 0.60,
+    BudgetLevel.COMPRESS: 0.80,
+    BudgetLevel.BLOCK: 0.90,
+    BudgetLevel.TERMINATE: 0.95,
+}
+
+
+class TokenBudget:
+    """Track token usage and enforce context limits."""
+
+    def __init__(self, max_tokens: int = 128000,
+                 thresholds: Optional[Dict[BudgetLevel, float]] = None):
+        self._max_tokens = max_tokens
+        self._thresholds = thresholds or THRESHOLDS
+        self._prompt_tokens = 0
+        self._completion_tokens = 0
+        self._tool_output_tokens = 0
+        self._history: List[Dict[str, Any]] = []
+
+    @property
+    def used_tokens(self) -> int:
+        return self._prompt_tokens + self._completion_tokens
+
+    @property
+    def remaining(self) -> int:
+        return max(0, self._max_tokens - self.used_tokens)
+
+    @property
+    def percentage(self) -> float:
+        if self._max_tokens == 0:
+            return 0
+        return self.used_tokens / self._max_tokens
+
+    def record_usage(self, prompt_tokens: int = 0, completion_tokens: int = 0,
+                     tool_output_tokens: int = 0):
+        """Record token usage from an API call."""
+        self._prompt_tokens += prompt_tokens
+        self._completion_tokens += completion_tokens
+        self._tool_output_tokens += tool_output_tokens
+        self._history.append({
+            "time": time.time(),
+            "prompt": prompt_tokens,
+            "completion": completion_tokens,
+            "tool_output": tool_output_tokens,
+            "total_used": self.used_tokens,
+        })
+
+    def check(self) -> BudgetStatus:
+        """Check current budget status and return appropriate actions."""
+        pct = self.percentage
+
+        if pct >= self._thresholds.get(BudgetLevel.TERMINATE, 0.95):
+            level = BudgetLevel.TERMINATE
+            msg = f"Context {pct:.0%} full. Session must terminate with summary."
+            actions = ["generate_summary", "terminate_session"]
+        elif pct >= self._thresholds.get(BudgetLevel.BLOCK, 0.90):
+            level = BudgetLevel.BLOCK
+            msg = f"Context {pct:.0%} full. Blocking verbose tool calls."
+            actions = ["block_verbose_tools", "force_wrap_up", "suggest_summary"]
+        elif pct >= self._thresholds.get(BudgetLevel.COMPRESS, 0.80):
+            level = BudgetLevel.COMPRESS
+            msg = f"Context {pct:.0%} full. Auto-compressing conversation."
+            actions = ["auto_compress", "drop_raw_tool_outputs", "suggest_summary"]
+        elif pct >= self._thresholds.get(BudgetLevel.WARNING, 0.60):
+            level = BudgetLevel.WARNING
+            msg = f"Context {pct:.0%} used. Consider summarizing."
+            actions = ["suggest_summary", "log_warning"]
+        else:
+            level = BudgetLevel.OK
+            msg = f"Context OK: {self.used_tokens}/{self._max_tokens} tokens ({pct:.0%})"
+            actions = []
+
+        return BudgetStatus(
+            level=level,
+            used_tokens=self.used_tokens,
+            max_tokens=self._max_tokens,
+            percentage=round(pct, 3),
+            remaining=self.remaining,
+            message=msg,
+            actions=actions,
+        )
+
+    def should_truncate_tool_output(self, estimated_tokens: int) -> bool:
+        """Check if a tool output should be truncated."""
+        if self.used_tokens + estimated_tokens > self._max_tokens * 0.95:
+            return True
+        return False
+
+    def get_truncation_budget(self) -> int:
+        """Get max tokens available for next tool output."""
+        budget = self.remaining - int(self._max_tokens * 0.05)  # Reserve 5%
+        return max(0, budget)
+
+    def reset(self):
+        """Reset budget for new session."""
+        self._prompt_tokens = 0
+        self._completion_tokens = 0
+        self._tool_output_tokens = 0
+        self._history.clear()
+
+    def get_report(self) -> Dict[str, Any]:
+        """Generate usage report."""
+        status = self.check()
+        return {
+            "status": status.level.value,
+            "used_tokens": self.used_tokens,
+            "max_tokens": self._max_tokens,
+            "remaining": self.remaining,
+            "percentage": status.percentage,
+            "prompt_tokens": self._prompt_tokens,
+            "completion_tokens": self._completion_tokens,
+            "tool_output_tokens": self._tool_output_tokens,
+            "message": status.message,
+            "actions": status.actions,
+        }

docs/cron-audit-890.md

@@ -1,50 +0,0 @@
-# Cron Audit & Cleanup
-
-Identifies and removes dead/stale cron jobs that waste scheduler cycles.
-
-## Problem
-
-Of 69 cron jobs, 9 had zero completions — running on schedule but never producing results. Each execution costs API tokens and scheduler cycles.
-
-## Dead Jobs Found (2026-04-14)
-
-| Job | Schedule | Completions |
-|-----|----------|-------------|
-| exp-swarm-pipeline | every 10 min | 0 |
-| exp-music-generator | every 2h | 0 |
-| exp-paper-citations | every 3h | 0 |
-| exp-gbrain-patterns | every 2h | 0 |
-| exp-infra-hardening | every 2h | 0 |
-| gemma4-multimodal-burn | every 1h | 0 |
-| morning-paper-report | daily | 0 |
-| overnight-collector | every 15 min | 0 |
-| morning-experiment-report | daily | 0 |
-
-## Usage
-
-```bash
-# Show dead jobs (zero completions)
-python3 scripts/cron_audit.py
-
-# Show all jobs with status
-python3 scripts/cron_audit.py --all
-
-# Find stale jobs (no runs in 7 days)
-python3 scripts/cron_audit.py --older-than 7
-
-# Disable dead jobs (pause them)
-python3 scripts/cron_audit.py --disable
-
-# Delete dead jobs permanently
-python3 scripts/cron_audit.py --delete
-
-# JSON output for automation
-python3 scripts/cron_audit.py --json
-```
-
-## Job Categories
-
-- **Healthy**: enabled, has completions, ran recently
-- **Dead**: enabled, zero completions, exists for N+ days
-- **Stale**: enabled, no successful run in N days
-- **Disabled**: paused or completed

scripts/cron_audit.py

@@ -1,279 +0,0 @@
-#!/usr/bin/env python3
-"""
-Cron Audit & Cleanup — find and remove dead/stale cron jobs.
-
-Identifies jobs that waste scheduler cycles:
-- Dead jobs: zero completions despite running for days
-- Stale jobs: no successful run in N days
-- Error jobs: high error ratio
-
-Usage:
-    # Show dead jobs (zero completions)
-    python3 scripts/cron_audit.py
-
-    # Show stale jobs (no runs in 7 days)
-    python3 scripts/cron_audit.py --older-than 7
-
-    # Show all jobs with status
-    python3 scripts/cron_audit.py --all
-
-    # Disable dead jobs (sets enabled=False, state=paused)
-    python3 scripts/cron_audit.py --disable
-
-    # Delete dead jobs permanently
-    python3 scripts/cron_audit.py --delete
-
-    # Custom threshold: dead = 0 completions after N days
-    python3 scripts/cron_audit.py --min-age 3
-
-    # JSON output
-    python3 scripts/cron_audit.py --json
-"""
-
-import argparse
-import json
-import os
-import sys
-from datetime import datetime, timedelta, timezone
-from pathlib import Path
-from typing import Any, Dict, List, Optional
-
-# hermes cron is a sibling module
-sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
-
-try:
-    from cron.jobs import (
-        list_jobs,
-        load_jobs,
-        save_jobs,
-        update_job,
-        remove_job,
-    )
-except ImportError:
-    # Fallback: direct file access
-    JOBS_FILE = Path.home() / ".hermes" / "cron" / "jobs.json"
-
-    def load_jobs() -> list:
-        if JOBS_FILE.exists():
-            return json.loads(JOBS_FILE.read_text())
-        return []
-
-    def save_jobs(jobs: list):
-        JOBS_FILE.parent.mkdir(parents=True, exist_ok=True)
-        JOBS_FILE.write_text(json.dumps(jobs, indent=2))
-
-    def list_jobs(include_disabled=False):
-        jobs = load_jobs()
-        if not include_disabled:
-            jobs = [j for j in jobs if j.get("enabled", True)]
-        return jobs
-
-    def update_job(job_id, updates):
-        jobs = load_jobs()
-        for job in jobs:
-            if job["id"] == job_id:
-                job.update(updates)
-                save_jobs(jobs)
-                return job
-        return None
-
-    def remove_job(job_id):
-        jobs = load_jobs()
-        original = len(jobs)
-        jobs = [j for j in jobs if j["id"] != job_id]
-        if len(jobs) < original:
-            save_jobs(jobs)
-            return True
-        return False
-
-
-# ── Analysis ──────────────────────────────────────────────────────────────
-
-def parse_schedule_display(job: dict) -> str:
-    """Get human-readable schedule from job."""
-    sched = job.get("schedule", {})
-    if isinstance(sched, dict):
-        return sched.get("display", job.get("schedule_display", "?"))
-    return str(sched)
-
-
-def get_last_run_age(job: dict) -> Optional[timedelta]:
-    """Get time since last run."""
-    last = job.get("last_run_at")
-    if not last:
-        return None
-    try:
-        if isinstance(last, str):
-            last_dt = datetime.fromisoformat(last.replace("Z", "+00:00"))
-        else:
-            return None
-        return datetime.now(timezone.utc) - last_dt
-    except (ValueError, TypeError):
-        return None
-
-
-def analyze_jobs(min_age_days: int = 0, stale_days: int = 0) -> Dict[str, List[dict]]:
-    """Analyze all jobs and categorize them.
-
-    Returns dict with keys: dead, stale, healthy, disabled, completed.
-    """
-    all_jobs = list_jobs(include_disabled=True)
-    now = datetime.now(timezone.utc)
-
-    result = {"dead": [], "stale": [], "healthy": [], "disabled": [], "completed": []}
-
-    for job in all_jobs:
-        job_id = job.get("id", "?")
-        name = job.get("name", job_id)
-        enabled = job.get("enabled", True)
-        state = job.get("state", "scheduled")
-        completed = job.get("repeat", {}).get("completed", 0)
-        schedule = parse_schedule_display(job)
-        last_run = job.get("last_run_at")
-        last_status = job.get("last_status", "never")
-        last_error = job.get("last_error")
-        created = job.get("created_at", "")
-
-        # Calculate age
-        age_days = 0
-        if created:
-            try:
-                created_dt = datetime.fromisoformat(created.replace("Z", "+00:00"))
-                age_days = (now - created_dt).days
-            except (ValueError, TypeError):
-                pass
-
-        last_age = get_last_run_age(job)
-
-        entry = {
-            "id": job_id,
-            "name": name,
-            "schedule": schedule,
-            "enabled": enabled,
-            "state": state,
-            "completed": completed,
-            "last_run_at": last_run,
-            "last_status": last_status,
-            "last_error": last_error,
-            "age_days": age_days,
-            "last_run_age_days": last_age.days if last_age else None,
-        }
-
-        if not enabled or state == "completed":
-            result["disabled"].append(entry)
-        elif completed == 0 and age_days >= min_age_days:
-            result["dead"].append(entry)
-        elif stale_days > 0 and last_age and last_age.days >= stale_days:
-            result["stale"].append(entry)
-        else:
-            result["healthy"].append(entry)
-
-    return result
-
-
-# ── Actions ───────────────────────────────────────────────────────────────
-
-def disable_jobs(jobs: List[dict]) -> int:
-    """Disable dead/stale jobs (pause them)."""
-    count = 0
-    for j in jobs:
-        result = update_job(j["id"], {"enabled": False, "state": "paused"})
-        if result:
-            count += 1
-            print(f"  DISABLED: {j['name']} ({j['schedule']})")
-    return count
-
-
-def delete_jobs(jobs: List[dict]) -> int:
-    """Permanently delete jobs."""
-    count = 0
-    for j in jobs:
-        if remove_job(j["id"]):
-            count += 1
-            print(f"  DELETED: {j['name']} ({j['schedule']})")
-    return count
-
-
-# ── Report ────────────────────────────────────────────────────────────────
-
-def print_table(jobs: List[dict], title: str):
-    """Print a table of jobs."""
-    if not jobs:
-        return
-    print(f"
-{title} ({len(jobs)}):")
-    print(f"  {'Name':<35} {'Schedule':<15} {'Completed':<10} {'Last Run':<15} {'Status'}")
-    print(f"  {'-'*35} {'-'*15} {'-'*10} {'-'*15} {'-'*10}")
-    for j in jobs:
-        last_run = "never"
-        if j["last_run_age_days"] is not None:
-            last_run = f"{j['last_run_age_days']}d ago"
-        elif j["last_run_at"]:
-            last_run = j["last_run_at"][:10]
-        status = j["last_status"] or "never"
-        print(f"  {j['name']:<35} {j['schedule']:<15} {j['completed']:<10} {last_run:<15} {status}")
-
-
-# ── CLI ───────────────────────────────────────────────────────────────────
-
-def main():
-    parser = argparse.ArgumentParser(description="Cron audit & cleanup")
-    parser.add_argument("--all", action="store_true", help="Show all jobs including healthy")
-    parser.add_argument("--older-than", type=int, default=0,
-                        help="Stale threshold in days (jobs with no runs in N days)")
-    parser.add_argument("--min-age", type=int, default=0,
-                        help="Minimum job age in days to be considered dead")
-    parser.add_argument("--disable", action="store_true", help="Disable dead jobs")
-    parser.add_argument("--delete", action="store_true", help="Delete dead jobs permanently")
-    parser.add_argument("--json", dest="json_output", action="store_true", help="JSON output")
-    args = parser.parse_args()
-
-    analysis = analyze_jobs(min_age_days=args.min_age, stale_days=args.older_than)
-
-    if args.json_output:
-        print(json.dumps(analysis, indent=2))
-        return
-
-    # Summary
-    total = sum(len(v) for v in analysis.values())
-    print(f"Cron Audit — {total} total jobs")
-    print(f"  Healthy:   {len(analysis['healthy'])}")
-    print(f"  Dead:      {len(analysis['dead'])}")
-    print(f"  Stale:     {len(analysis['stale'])}")
-    print(f"  Disabled:  {len(analysis['disabled'])}")
-    print(f"  Completed: {len(analysis['completed'])}")
-
-    if args.all:
-        print_table(analysis["healthy"], "HEALTHY")
-    print_table(analysis["dead"], "DEAD (zero completions)")
-    print_table(analysis["stale"], "STALE (no recent runs)")
-
-    if not args.disable and not args.delete:
-        if analysis["dead"] or analysis["stale"]:
-            print(f"
-To clean up: --disable (pause) or --delete (permanent)")
-        return
-
-    targets = analysis["dead"] + analysis["stale"]
-    if not targets:
-        print("
-Nothing to clean up.")
-        return
-
-    if args.delete:
-        confirm = input(f"
-Delete {len(targets)} jobs permanently? [y/N] ")
-        if confirm.lower() != "y":
-            print("Aborted.")
-            return
-        count = delete_jobs(targets)
-        print(f"
-Deleted {count} jobs.")
-    elif args.disable:
-        count = disable_jobs(targets)
-        print(f"
-Disabled {count} jobs.")
-
-
-if __name__ == "__main__":
-    main()

tests/test_cron_audit.py

@@ -1,143 +0,0 @@
-"""Tests for cron audit script."""
-
-import json
-import os
-import sys
-from datetime import datetime, timedelta, timezone
-from pathlib import Path
-from unittest.mock import MagicMock, patch
-
-import pytest
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "scripts"))
-
-
-# ---------------------------------------------------------------------------
-# Test data
-# ---------------------------------------------------------------------------
-
-def _make_job(name, job_id=None, completed=0, enabled=True, state="scheduled",
-              last_run=None, last_status=None, created=None, schedule="every 1h"):
-    return {
-        "id": job_id or f"job-{name}",
-        "name": name,
-        "enabled": enabled,
-        "state": state,
-        "schedule": {"display": schedule} if isinstance(schedule, str) else schedule,
-        "repeat": {"completed": completed},
-        "last_run_at": last_run,
-        "last_status": last_status,
-        "created_at": created or datetime.now(timezone.utc).isoformat(),
-    }
-
-
-# ---------------------------------------------------------------------------
-# analyze_jobs
-# ---------------------------------------------------------------------------
-
-class TestAnalyzeJobs:
-    @patch("cron_audit.list_jobs")
-    def test_dead_job_detected(self, mock_list):
-        """Job with 0 completions and old creation date is dead."""
-        old_date = (datetime.now(timezone.utc) - timedelta(days=5)).isoformat()
-        mock_list.return_value = [
-            _make_job("dead-job", completed=0, created=old_date),
-        ]
-        from cron_audit import analyze_jobs
-        result = analyze_jobs(min_age_days=1)
-        assert len(result["dead"]) == 1
-        assert result["dead"][0]["name"] == "dead-job"
-
-    @patch("cron_audit.list_jobs")
-    def test_healthy_job_not_dead(self, mock_list):
-        """Job with completions is healthy."""
-        mock_list.return_value = [
-            _make_job("good-job", completed=42, last_run=datetime.now(timezone.utc).isoformat()),
-        ]
-        from cron_audit import analyze_jobs
-        result = analyze_jobs()
-        assert len(result["dead"]) == 0
-        assert len(result["healthy"]) == 1
-
-    @patch("cron_audit.list_jobs")
-    def test_disabled_job_categorized(self, mock_list):
-        """Disabled job goes to disabled category."""
-        mock_list.return_value = [
-            _make_job("paused-job", enabled=False, state="paused"),
-        ]
-        from cron_audit import analyze_jobs
-        result = analyze_jobs()
-        assert len(result["disabled"]) == 1
-
-    @patch("cron_audit.list_jobs")
-    def test_stale_job_detected(self, mock_list):
-        """Job with last run > N days ago is stale."""
-        old_run = (datetime.now(timezone.utc) - timedelta(days=10)).isoformat()
-        mock_list.return_value = [
-            _make_job("stale-job", completed=5, last_run=old_run),
-        ]
-        from cron_audit import analyze_jobs
-        result = analyze_jobs(stale_days=7)
-        assert len(result["stale"]) == 1
-
-    @patch("cron_audit.list_jobs")
-    def test_completed_job_disabled(self, mock_list):
-        """Job with state=completed goes to disabled."""
-        mock_list.return_value = [
-            _make_job("done-job", completed=100, state="completed", enabled=False),
-        ]
-        from cron_audit import analyze_jobs
-        result = analyze_jobs()
-        assert len(result["disabled"]) == 1
-
-    @patch("cron_audit.list_jobs")
-    def test_empty_jobs(self, mock_list):
-        """No jobs returns empty categories."""
-        mock_list.return_value = []
-        from cron_audit import analyze_jobs
-        result = analyze_jobs()
-        for v in result.values():
-            assert len(v) == 0
-
-
-# ---------------------------------------------------------------------------
-# get_last_run_age
-# ---------------------------------------------------------------------------
-
-class TestGetLastRunAge:
-    def test_returns_timedelta(self):
-        from cron_audit import get_last_run_age
-        recent = (datetime.now(timezone.utc) - timedelta(hours=2)).isoformat()
-        age = get_last_run_age({"last_run_at": recent})
-        assert age is not None
-        assert age.days == 0
-        assert age.seconds >= 7000  # ~2 hours
-
-    def test_returns_none_for_no_run(self):
-        from cron_audit import get_last_run_age
-        assert get_last_run_age({"last_run_at": None}) is None
-        assert get_last_run_age({}) is None
-
-
-# ---------------------------------------------------------------------------
-# disable_jobs / delete_jobs
-# ---------------------------------------------------------------------------
-
-class TestJobActions:
-    @patch("cron_audit.update_job")
-    def test_disable_calls_update(self, mock_update):
-        mock_update.return_value = {"id": "x"}
-        from cron_audit import disable_jobs
-        jobs = [{"id": "j1", "name": "test", "schedule": "1h"}]
-        count = disable_jobs(jobs)
-        assert count == 1
-        mock_update.assert_called_once_with("j1", {"enabled": False, "state": "paused"})
-
-    @patch("cron_audit.remove_job")
-    def test_delete_calls_remove(self, mock_remove):
-        mock_remove.return_value = True
-        from cron_audit import delete_jobs
-        jobs = [{"id": "j1", "name": "test", "schedule": "1h"}]
-        count = delete_jobs(jobs)
-        assert count == 1
-        mock_remove.assert_called_once_with("j1")

tools/skill_manager_tool.py

@@ -44,6 +44,34 @@ from typing import Dict, Any, Optional, Tuple
 
 logger = logging.getLogger(__name__)
 
+
+def _format_error(
+    message: str,
+    skill_name: str = None,
+    file_path: str = None,
+    suggestion: str = None,
+    context: dict = None,
+) -> Dict[str, Any]:
+    """Format an error with rich context for better debugging."""
+    parts = [message]
+    if skill_name:
+        parts.append(f"Skill: {skill_name}")
+    if file_path:
+        parts.append(f"File: {file_path}")
+    if suggestion:
+        parts.append(f"Suggestion: {suggestion}")
+    if context:
+        for key, value in context.items():
+            parts.append(f"{key}: {value}")
+    return {
+        "success": False,
+        "error": " | ".join(parts),
+        "skill_name": skill_name,
+        "file_path": file_path,
+        "suggestion": suggestion,
+    }
+
+
 # Import security scanner — agent-created skills get the same scrutiny as
 # community hub installs.
 try: