superpowers

Release v6.4.2 with leaner plans

This release removes orphaned Codex hooks and documentation references, refactors skill documentation to reduce duplication and improve clarity, and enhances the writing-good-tests skill with falsifiability discipline and mutation testing guidance. The finishing-a-development-branch skill is streamlined to remove the discard work offer from completion menus and made forge-agnostic, while writing-good-tests is reframed as a positive catalog of rules with new testing anti-pattern guidance.

Release v6.4.1 with Codex fixes

Fixed Codex SessionStart hook auto-discovery by declaring an empty hooks object in the plugin manifest, removed orphaned Codex-specific hook code, refactored skill documentation to remove duplication and improve clarity, and restored Gemini CLI support. This release includes improvements to the finishing-a-development-branch skill workflow and consolidates testing best practices documentation.

Release v6.3.0 with CLI and agent support

Version 6.3.0 adds support for Devin CLI and Hermes Agent, implements a three-path brainstorming router, and improves efficiency across SDD and Codex systems. Key fixes include suppressing unwanted SessionStart hook auto-discovery, refactoring skill documentation to reduce duplication, and strengthening test writing practices.

Removed hiring section from README

Removed the "We're Hiring" section from the project README since the community engineer position now has a candidate in the trial phase.

Version 6.2.0 release

Superpowers v6.2.0 ships structural improvements to Subagent-Driven Development (plan-scoped workspaces and resume-based fix loops), a comprehensive skills library compression campaign, Windows SessionStart hook fixes, restored Gemini CLI support, and fixes for test discovery, cross-platform packaging, and documentation. The update spans all seven plugin manifests and includes detailed release notes covering lifecycle improvements, skill reorganization, and bug fixes.

Fix test pattern matching for top-level files

Fixed the find-polluter debugging script to correctly match test files at the top level of directories (e.g., src/top.test.ts) and accept patterns with or without a leading ./ prefix. The script now searches using multiple pattern variations to ensure no test files are skipped.

Fix test file discovery in debug script

Corrected a script that searches for test files to properly match paths that start with './'. The script was failing to find any test files because it wasn't accounting for the './' prefix that the find command outputs, causing it to incorrectly report results.

Remove broken documentation link

Fixed a broken cross-reference in the Antigravity tools documentation that pointed to a section that no longer exists. The relevant guidance about subagent types (`self` vs `research`) remains in the same table cell, so no information was lost.

Restructured subagent development lifecycle

Updated the subagent-driven development skill documentation with a new resume-based fix loop, a five-round breaker mechanism to prevent infinite cycles, and a rationalization table for clearer process guidance.

Refined fix round workflows

Updated guidance for handling task review findings and fix cycles in subagent-driven development. The implementer now receives findings via resumption, runs covering tests, and appends a detailed fix report before replying with a status update. Clarified when to close reviewer subagents (after review returns) and how to manage implementer subagents through fix loops.