Skip to content

Security: mishanefedov/agentwatch

Security

SECURITY.md

Security Policy

agentwatch is a local observability tool. It reads files from your home directory (Claude Code logs, OpenClaw workspaces, Cursor config, workspace filesystem). It never sends data over the network.

If you find a vulnerability, please report it privately before opening a public issue.

Reporting

Open a private Security Advisory on GitHub — that keeps the report confidential until a fix is out.

Please include:

  • Version affected (output of agentwatch --version or package version)
  • OS + Node version
  • Minimal reproduction steps
  • Observed impact

I'll respond within 7 days.

Scope

In scope:

  • Arbitrary file read outside the documented watched paths
  • Exfiltration of file contents or metadata to the network
  • Command injection through watched file names or log contents
  • Prototype pollution or similar JavaScript-ecosystem vulns in the parser
  • Denial of service against the running TUI from crafted log lines
  • Supply-chain issues in the npm package

Out of scope:

  • Social-engineering the maintainer
  • Bugs in Claude Code / Cursor / OpenClaw themselves
  • Vulnerabilities in the terminal emulator, OS clipboard tools, notification daemons, or other OS-provided infrastructure agentwatch shells out to

What agentwatch reads

Full list documented in the README, but for quick reference:

  • ~/.claude/projects/**/*.jsonl (Claude Code session transcripts + subagents)
  • ~/.claude/settings.json (Claude permissions)
  • ~/.openclaw/agents/*/sessions/*.jsonl (OpenClaw sub-agent sessions)
  • ~/.openclaw/logs/config-audit.jsonl (OpenClaw config audit trail)
  • ~/.openclaw/openclaw.json (OpenClaw agent roster)
  • ~/.cursor/mcp.json, cli-config.json, ide_state.json (Cursor state)
  • Project-level .cursorrules files under $WORKSPACE_ROOT
  • The $WORKSPACE_ROOT tree (defaults to ~/IdeaProjects) for file-change events

agentwatch writes only to the terminal and to the system clipboard (on explicit y key press). No files are created or modified outside of its own repo during development.

What agentwatch does NOT do

  • No outbound HTTP / DNS / network calls (verify with lsof -i -p $(pgrep -n agentwatch))
  • No telemetry. Not opt-in, not opt-out — it's just absent.
  • No account, no cloud, no sign-in.
  • No data leaves your machine.

There aren't any published security advisories