Argus v0.2.0

Security hardening release: Unicode-normalized runtime prompt blocking, deny-by-default tools, OpenAI Responses function-call enforcement, bounded repository ingress, fail-closed audit recovery, hardened containers, explicit baseline decisions, GitHub SARIF upload, and refreshed real MCP evidence.

Argus v0.1.0

Changelog

All notable changes to Argus are documented here. The project follows a
lightweight Keep a Changelog style; releases are versioned in pyproject.toml.

[Unreleased]

• Added baseline/diff gating so CI can fail on new or more-severe findings while
  existing findings are being remediated.
• Added a cross-platform Python portfolio demo and an intentionally vulnerable
  example agent.
• Added a Mermaid architecture diagram, sanitized real-MCP evidence record,
  repository security policy, contribution guide, and license.
• Added CI compatibility and dependency supply-chain checks.

[0.1.0]

• Added static agent, MCP, and skill analysis.
• Added bounded LLM probes and read-only stdio/Streamable HTTP MCP discovery.
• Added JSON, Markdown, and SARIF reporting.
• Added the optional runtime policy gateway with metrics and hash-chained audit
  events.