Skip to content

Security: Agentic-CLI/agentic-cli

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please do not open a public issue for security reports. Email security@agentic-cli.com with details and reproduction steps. We aim to acknowledge within 3 business days and will coordinate a fix and disclosure timeline with you.

Security posture of the tool itself

Agentic CLI is designed to be safe to run inside any repository:

  • No runtime dependencies. Nothing is pulled from the network at runtime; the supply-chain surface is the standard library plus (optionally) PyYAML.
  • Local-first. The trust ledger and relay items are plain files under .agentic/. Nothing leaves your machine unless you explicitly wire a cloud sync.
  • No secrets handled or stored. The tool records provenance metadata (paths, decisions, run IDs) — not file contents, prompts, or credentials.
  • Tamper-evident ledger. Entries are chained with SHA-256 so that editing or removing any past entry is detectable (agentic doctor verifies the chain).

    Note: the hash chain is integrity-evident, not a cryptographic signature. Signed provenance (ed25519) is on the roadmap; until then, treat the ledger as tamper-evident, not tamper-proof.

  • Runs at the harness's own hook boundary — it does not execute or modify agent code, and blocking a tool call fails safe (the action simply does not proceed).

Supported versions

This project is pre-1.0 (alpha). Security fixes are applied to main.

There aren't any published security advisories