Trending on GitHub

Mozilla has revoked its Firefox code signing key after an unencrypted copy was discovered in a GitHub repository. Internal audit logs indicate no unauthorized access occurred during the incident. The organization is now updating its release verification processes to prevent similar exposures.

• Firefox signing key revoked immediately after unencrypted copy surfaced on GitHub
• Audit logs confirmed no unexpected visitors or data exfiltration during the event
• Release verification workflows require updates to mitigate future signing key risks

The msitarzewski/agency-agents repository provides a curated collection of specialized AI agents, each designed with specific personas and workflows for tasks ranging from frontend development to community management. A native desktop application is available for macOS, Linux, and Windows, allowing users to browse the roster and install agents directly into IDEs like Claude Code, Cursor, and Codex with a single click. The tool auto-updates and eliminates the need for manual cloning or scripting to integrate these expert profiles.

• Native app simplifies installation of specialized AI agents into major IDEs like Cursor and Claude Code.
• Agents come pre-configured with distinct personas, processes, and deliverables for specific roles.
• Auto-update feature ensures agent definitions stay current without manual intervention.
• Cross-platform support covers macOS, Linux, and Windows for broad engineer accessibility.

OpenMontage is an open-source platform that aggregates over 100 tools and 700 agent skills to automate video production. It enables multiple AI models like Claude and ChatGPT to collaborate within a single conversation to execute complex video pipelines. The system integrates with coding assistants to turn them into full video production studios without requiring local setup.

• Aggregates 12 pipelines and 100+ tools for automated video creation
• Enables multi-agent collaboration between models like Claude and ChatGPT
• Integrates with AI coding assistants to expand their utility
• Runs in the cloud with zero local setup required

Code-Graph-RAG parses multi-language codebases using Tree-sitter to construct a knowledge graph stored in Memgraph. It enables querying, editing, and optimizing code through natural language interfaces by mapping repository structure into a unified schema. The system supports structural search and replace via ast-grep agent tools and now includes pluggable Ruby support using YAML-based patterns.

• Builds a unified knowledge graph across mixed-language monorepos using Tree-sitter and Memgraph.
• Enables natural language querying and editing of code structures without hand-written parsers for every language.
• Integrates ast-grep for structural search and replace, exposing these capabilities as agent tools.
• Supports pluggable language extensions, such as the new Ruby tier defined via simple YAML pattern files.