Перейти к содержимому
TakoAPI
Назад к навыкам

agentguard

от manas-io-ai

DevOps & Cloud
2.8kЧисло загрузок по данным clawskills.sh
3Звёзды
0Просмотры
0 лайковВойдите, чтобы поставить лайк навыку
—Ещё нет оценок
Ваша оценка

Кратко

**Category:** Security & Monitoring.

Описание

Что делает этот навык

Monitors agent file access, outbound API calls, and external communications to detect suspicious behavior. Generates security reports and sends alerts with configurable severity thresholds. All processing runs locally with no external data transmission.

Runs entirely locally with no external data transmission, so sensitive operational data stays on the machine while still providing full audit trails and ML-lite anomaly detection.

Когда его использовать

  • Detecting when an agent reads credential files like .env or API key files
  • Auditing outbound HTTP calls made by a running agent during a session
  • Generating a daily security summary of agent activity and anomalies
  • Investigating file access patterns over a specific time window after a suspected issue
  • Flagging bulk file reads that match data exfiltration patterns

Пример сценария

ВВОД

User asks: 'What has my agent been doing with files in the last 2 hours?'

АГЕНТ
  1. 1

    Run agentguard start to confirm monitoring is active

  2. 2

    Run agentguard investigate --timerange 'last 2 hours' --type file_access

  3. 3

    Review returned events grouped by severity level

  4. 4

    Run agentguard alerts --last 24h --severity high to check for flagged incidents

  5. 5

    Run agentguard report --now for a full activity summary

ВЫВОД

A breakdown of file access events with severity labels, any triggered alerts, and a list of accessed resource paths from the past two hours

Установка

clawhub install manas-io-ai/agentguard

Выполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.

Ссылки