Prompt file imported from 0x0funky/Agentinel (
.codex/prompts/agentinel.md). Copyright stays with the author.
Use the Agentinel CLI to inspect this machine's live resources and advise the user.
Agentinel is a zero-overhead CLI (~12 MB, no WebView) that snapshots every process, classifies AI-agent CLIs (Claude Code, Codex, Gemini, OpenCode, MCP & dev servers, Ollama…), and flags leaks / zombies / duplicate sessions / heavy disk.
Run from the repo root (or anywhere agent_monitor/ is importable). Prefer JSON:
agentinel scan --json # whole-system snapshot (parse this)
agentinel scan --agents-only --json # AI agents only
agentinel advise --engine codex # local AI writes markdown advice
agentinel doctor # which AI CLIs are installed
# (or: python -m agent_monitor <subcommand> …)scan --json returns { system, processes[], findings[] }:
- system: ram_used_gb / ram_total_gb / ram_pct, cpu_pct, swap_*
- each process: pid, label, category, project, mem_mb, cpu, age_hr, idle_min, parent_name, parent_alive, ports[], children, cmdline
- findings: rule-engine anomalies — { severity, title, detail, pids, action }
Then: summarize the biggest RAM users / suspected leaks / idle zombies / duplicate agent sessions, citing evidence (mem, idle_min, parent_alive=false, ports), and recommend keep / end / watch for each.
IMPORTANT — these commands are READ-ONLY. There is no CLI kill or cleanup; never
invent one. Destructive actions (kill / suspend / disk cleanup) only happen in the
GUI (python agentinel.py) after the user confirms. Don't run anything destructive
without explicit user confirmation.