gsd-build/get-shit-done v1.27.0

on GitHub

[1.27.0] - 2026-03-20

Added

• Advisor mode — Research-backed discussion with parallel agents evaluating gray areas before you decide
• Multi-repo workspace support — Auto-detection and project root resolution for monorepos and multi-repo setups
• Cursor CLI runtime support — Full installation and command conversion for Cursor
• /gsd:fast command — Trivial inline tasks that skip planning entirely
• /gsd:review command — Cross-AI peer review of current phase or branch
• /gsd:plant-seed command — Backlog parking lot for ideas and persistent context threads
• /gsd:pr-branch command — Clean PR branches filtering .planning/ commits
• /gsd:audit-uat command — Verification debt tracking across phases
• --analyze flag for discuss-phase — Trade-off analysis during discussion
• research_before_questions config option — Run research before discussion questions instead of after
• Ticket-based phase identifiers — Support for team workflows using ticket IDs
• Worktree-aware .planning/ resolution — File locking for safe parallel access
• Discussion audit trail — Auto-generated DISCUSSION-LOG.md during discuss-phase
• Context window size awareness — Optimized behavior for 1M+ context models
• Exa and Firecrawl MCP support — Additional research tools for research agents
• Decision IDs — Discuss-to-plan traceability via decision identifiers
• Stub detection — Verifier and executor detect incomplete implementations
• Security hardening — Prompt injection guards, path traversal prevention, input validation

Changed

• CI matrix updated to Node 20, 22, 24 — dropped EOL Node 18
• GitHub Actions upgraded for Node 24 compatibility
• Consolidated planningPaths() helper — eliminated 34 inline path constructions
• Materialize full config on new-project initialization

Fixed

• Codex config.toml corruption from non-boolean [features] keys
• Universal agent name replacement for non-Claude runtimes
• --no-verify support for parallel executor commits
• ROADMAP fallback for plan-phase, execute-phase, and verify-work
• Copilot sequential fallback and spot-check completion detection
• Cursor: preserve slash-prefixed commands and unquoted skill names
• Semver 3+ segment parsing and CRLF frontmatter corruption recovery
• STATE.md parsing fixes (compound Plan field, progress tables, lifecycle extraction)
• Windows HOME sandboxing for tests
• Various hook, config, and cross-platform fixes

Install/upgrade: npx get-shit-done-cc@latest