health
The health command scans for stale or zombie Claude Code and Codex sessions across your system, displaying session details like PID, elapsed time, and flags. Use this to identify long-running or abandoned sessions that may be consuming resources, then safely terminate selected sessions with explicit confirmation and dry-run previews before any changes take effect.
mkdir -p ~/.claude/commands && curl -fsSL https://raw.githubusercontent.com/alexgreensh/token-optimizer/HEAD/commands/health.md -o ~/.claude/commands/health.mdhealth.md
# Session Health Check
Run a session health check and help the user manage running sessions safely.
## Steps
1. Resolve measure.py path:
```bash
RUNTIME="${TOKEN_OPTIMIZER_RUNTIME:-}"
if [ -z "$RUNTIME" ]; then
# Env signals are authoritative and checked before directory heuristics: a host
# with BOTH ~/.codex and ~/.config/opencode (running OpenCode) must resolve to
# opencode, not codex, so the tool never reaches into ~/.claude (issue #57).
if [ -n "$CLAUDE_PLUGIN_ROOT" ] || [ -n "$CLAUDE_PLUGIN_DATA" ]; then
RUNTIME="claude"
elif [ -n "$OPENCODE" ] || [ -n "$OPENCODE_BIN" ] || [ -n "$OPENCODE_CONFIG_DIR" ] || [ -n "$OPENCODE_CONFIG" ]; then
RUNTIME="opencode"
elif [ -n "$CODEX_HOME" ]; then
RUNTIME="codex"
elif [ -n "$CLAUDECODE" ] || [ -n "$CLAUDE_CODE_ENTRYPOINT" ] || [ -n "$CLAUDE_CODE_SESSION_ID" ]; then
RUNTIME="claude"
elif [ -d "$HOME/.config/opencode" ] && [ ! -d "$HOME/.codex" ]; then
RUNTIME="opencode"
elif [ -d "$HOME/.codex" ]; then
RUNTIME="codex"
else
RUNTIME="claude"
fi
fi
# Resolve measure.py to the NEWEST installed copy across channels so a stale
# plugin-cache copy never shadows a fresh install (issue #57). find -L follows the
# install.sh symlink under ~/.claude/skills; cd -P resolves it before reading each
# copy's plugin.json for its version. find (not bare globs) never errors under zsh.
MEASURE_PY=""; TO_LAUNCHER=""; _best_ver=""
while IFS= read -r _cand; do
[ -f "$_cand" ] || continue
_root="$(cd -P -- "$(dirname -- "$_cand")/../../.." 2>/dev/null && pwd)"
_ver="$(sed -n 's/.*"version"[[:space:]]*:[[:space:]]*"\([^"]*\)".*/\1/p' "$_root/.claude-plugin/plugin.json" 2>/dev/null | head -1)"
[ -n "$_ver" ] || _ver="0.0.0"
if [ -z "$_best_ver" ] || [ "$(printf '%s\n%s\n' "$_ver" "$_best_ver" | sort -t. -k1,1n -k2,2n -k3,3n -k4,4n | tail -n1)" = "$_ver" ]; then
_best_ver="$_ver"; MEASURE_PY="$_cand"; TO_LAUNCHER="$_root/hooks/python-launcher.sh"
fi
done <<EOF
$(find -L "$HOME/.claude/skills" "$HOME/.claude/plugins/cache" "$HOME/.claude/token-optimizer" "$HOME/.codex/skills" "$HOME/.codex/plugins/cache" "$HOME/.config/opencode/plugins/cache" "$HOME/.config/opencode/plugins" -type f -name measure.py -path '*token-optimizer*/scripts/measure.py' 2>/dev/null)
EOF
if [ -z "$MEASURE_PY" ]; then echo "[Error] measure.py not found. Is Token Optimizer installed?"; exit 1; fi
# python-launcher.sh sits at the plugin root beside skills/. Routing EVERY
# runtime through it keeps Windows invocations flash-free (#107): bare
# `python3` is a console-subsystem spawn on a Codex/OpenCode Windows host,
# while the launcher swaps to the GUI-subsystem pythonw.exe. On POSIX the
# launcher resolves the same python3 it always did.
[ -f "$TO_LAUNCHER" ] || TO_LAUNCHER=""
export TOKEN_OPTIMIZER_RUNTIME="$RUNTIME"
```
2. Run (use the resolved `$RUNTIME` — never hardcode a runtime; under OpenCode this
keeps the session scan scoped to OpenCode and never reaches into `~/.claude`):
- Claude Code plugin: `bash "$CLAUDE_PLUGIN_ROOT/hooks/python-launcher.sh" $MEASURE_PY health`
- Codex / OpenCode / standalone: `TOKEN_OPTIMIZER_RUNTIME="$RUNTIME" bash "$TO_LAUNCHER" "$MEASURE_PY" health`
(only if `$TO_LAUNCHER` is empty, fall back to `TOKEN_OPTIMIZER_RUNTIME="$RUNTIME" python3 "$MEASURE_PY" health` — the bare-python3 form flashes a console window on Windows, #107)
3. Present results clearly. For each session show: PID, elapsed time, version, and flags (STALE >24h, ZOMBIE >48h, OUTDATED, HEADLESS, TERMINAL).
4. If ANY sessions are flagged STALE or ZOMBIE, ask the user:
"I found N session(s) that look stale. Want me to show details so you can decide which to terminate?"
5. **CRITICAL SAFETY RULES — follow these exactly:**
- NEVER auto-kill anything. Always ask first and get explicit confirmation.
- HEADLESS sessions might be intentional background processes (cron agents, heartbeat monitors, scheduled tasks). Always warn: "This session is headless, it might be a background agent running on purpose. Are you sure you want to terminate it?"
- Let the user pick specific PIDs to terminate, or offer "terminate all ZOMBIE-flagged sessions" as a batch option.
- Always run a dry-run first to preview what would be terminated, then ask for confirmation before running without `--dry-run`.
- Claude Code plugin dry-run: `bash "$CLAUDE_PLUGIN_ROOT/hooks/python-launcher.sh" $MEASURE_PY kill-stale --dry-run`
- Codex / OpenCode / standalone dry-run: `TOKEN_OPTIMIZER_RUNTIME="$RUNTIME" bash "$TO_LAUNCHER" "$MEASURE_PY" kill-stale --dry-run`
(only if `$TO_LAUNCHER` is empty, fall back to `TOKEN_OPTIMIZER_RUNTIME="$RUNTIME" python3 "$MEASURE_PY" kill-stale --dry-run`)
- If the user says "kill all" or similar, still show the dry-run preview and confirm. No silent kills.
6. If no stale or zombie sessions found, say: "All sessions look healthy. Your oldest is Xh old."Quick 10-second context health check with quality score and top issues
Audit a Claude Code or Codex setup for context-window waste, then fix it and measure the savings. Use when context feels tight.
Cross-system agent token/cost audit (Claude Code, Codex, OpenClaw, Hermes, OpenCode): idle burns, model misrouting, config bloat, with dollar savings.
Plan a token-efficient Claude Code or Codex setup, or get a quick health check. Coaching, not the full audit (use token-optimizer for that).
Open the Token Optimizer dashboard in your browser (context usage, quality, savings). Use to view the dashboard.
Pull a prior session's checkpoint on demand when the user is continuing prior work. Returns fenced, source-labeled, scrubbed recovery context. Do NOT call on a fresh, unrelated task.