Repository navigation
Description
v0.5.0 — Last Breaking Change Before v1.0
Timeline: Feb 17 - Mar 16, 2026 (4 weeks)
Release Date: March 16, 2026
Status: In Progress (Week 1 Day 2)
Breaking Change
Primary: Consolidate to single .squad/ directory (#69)
.ai-team/→.squad/.ai-team-templates/→.squad/templates/(nested)- ~1,672 occurrences across 130+ files
- Mandatory migration via
squad upgrade --migrate-directory - Backward compatibility window: v0.5.0 → v1.0.0
- Full legacy removal: v1.0.0
Rationale: One folder to find, one folder to .gitignore, one folder to understand. No confusion about where Squad state lives.
MUST Ship Issues
| Issue | Title | Owner | Estimate |
|---|---|---|---|
| #69 | Consolidate to .squad/ (directory + templates) | Fenster | 85h |
| #76 | Refactor squad.agent.md for GHE 30KB limit | Verbal | 24h |
| #86 | Squad undid uncommitted changes (HIGH SEVERITY) | Fenster + Hockney | 6-12h |
| #71 | Cleanup label workflows | Fenster | 18h |
| #84 | Add timestamps to session logs | Fenster | 12h |
| #62 | CI/CD integration patterns | Kobayashi | 28h |
Total Core: ~173-179 squad-hours
Bundled Enhancements
| PR | Title | Status | Shipped |
|---|---|---|---|
| #80 | Add llms.txt outputs to docs build | ✅ Merged to dev | Feb 17 |
| #89 | Fix MCP Discord notifications setup docs | ✅ Merged to dev | Feb 17 |
Both ride with v0.5.0 release March 16.
Launch Features
Insider Program (Week 1 Priority)
Binary access model: Insider (continuous dev/nightly) or Release (stable).
WHY FIRST: Testing infrastructure enables incremental validation throughout sprint. Contributors test #69, #76, #86 fixes as they merge to dev, not waiting until Week 3 beta.
Installation:
- Regular:
npx github:bradygaster/squad - Insider:
npx github:bradygaster/squad#insider - Upgrade:
squad upgradesupports regular → insider switching
Architecture:
- Separate
insiderbranch ("dev minus the team state") - Isolated state:
.squad-insider/directory - Version:
v0.5.0-insider+{commit}for bug reporting - Honor-system access + public list in CONTRIBUTORS.md
No rings, no caps: Everyone who contributes meaningfully can be an insider.
Recognition: [INSIDER] badge in CONTRIBUTORS.md, Discord #squad-insiders channel, blog credits, release notes.
Responsibilities: Test within 48-72h, file detailed bugs, validate exit criteria (optional).
Supporting Initiatives
Blogging (McManus)
Target: 4 blog posts during v0.5.0 cycle
- Week 1 (Feb 20-21): v0.5.0 kickoff — insider program opens, migration preview, Squad undid its own uncommited changes from a previous session #86 findings
- Week 2 (Feb 27-28): Under the hood — squad.agent.md split (Squads in GitHub Enterprise Copilot- Invalid config "Prompt" exceeds max length 30000 #76), single-folder rationale
- Week 3 (Mar 6-7): Insider testing — early signals, migration feedback
- Week 4 (Mar 16): v0.5.0 ships — migration guide, insider invitation confirmation, Squad undid its own uncommited changes from a previous session #86 fix announcement
Architectural Analysis (#86 & #87)
Decision (Feb 16, comment #3911872475):
- Squad undid its own uncommited changes from a previous session #86 (data loss bug): STAYS in v0.5.0. HIGH SEVERITY, trust-destroying. Week 1 investigation ✅ COMPLETE (prompt-only fix, 3-4h). Don't ship v0.5.0 until fix validated across 4 test scenarios.
- squad- workflows seem to assume a particular project type? #87 (workflow npm assumptions): DEFERRED to v0.6.0. Polish feature, user workaround exists (delete 3 files), not a blocker, scope protection paramount.
Deferred to v0.6.0
| Issue | Title | Rationale |
|---|---|---|
| #63 | Review label taxonomy | Post-migration polish |
| #82 | Session log compression | Performance optimization |
| #85 | Skill confidence auto-bump | Intelligence enhancement |
| #36 | JetBrains + GitHub.com research | After VS Code parity (#32-35) |
| #79 | Historical decision search | Discovery feature |
| #87 | Workflow project type detection | Template polish, not blocker |
| #75 | Agent spawning metrics | Observability |
Mandatory Beta Program
Cohort size: 5-10 users
Duration: Week 3-4 (Mar 3-16)
Target: Power users, existing contributors, early adopters
Exit criteria (all required):
- Migration succeeds on 5 real repos (with existing
.ai-team/state) - Zero state corruption reports
- Upgrade path tested (v0.4.1 → v0.5.0)
- Workflows pass on migrated repos
- Documentation validates (migration guide, troubleshooting)
- Squad undid its own uncommited changes from a previous session #86 fix validated (no more uncommitted work loss)
- Performance acceptable (no >10% overhead vs v0.4.1)
Timeline
Week 1 (Feb 17-23): Insider Program + Critical Investigation
PRIMARY: Insider Program infrastructure (enables incremental testing)
- Set up
insiderbranch + CI/CD automation - Document install paths (regular vs insider)
- CONTRIBUTORS.md badge system
- Discord #squad-insiders channel setup
- First insider publish (v0.5.0-insider+{commit})
SECONDARY: Critical bug investigation
- Squad undid its own uncommited changes from a previous session #86 data loss bug ✅ COMPLETE (Fenster: reproduced, root cause identified, prompt-only fix designed)
- Rename .ai-team/ to .squad/ with backward-compatible migration #69 audit ✅ COMPLETE (Fenster: 1,672 references across 130+ files, 3-PR plan)
- Squads in GitHub Enterprise Copilot- Invalid config "Prompt" exceeds max length 30000 #76 architecture ✅ COMPLETE (Verbal: 108KB → <30KB split, lib/ design)
Blog post #1 (Feb 20-21): v0.5.0 kickoff — insider program opens NOW
Week 2 (Feb 24 - Mar 2): Implementation Wave 1
- Squad undid its own uncommited changes from a previous session #86 fix implementation (Verbal: prompt updates, Hockney: test scenarios) — 6h
- Rename .ai-team/ to .squad/ with backward-compatible migration #69 PR feat: worktree awareness, Scribe auto-commit, and decision consolidation #1: CLI foundation + migration command — 8h
- Squads in GitHub Enterprise Copilot- Invalid config "Prompt" exceeds max length 30000 #76 PR feat: worktree awareness, Scribe auto-commit, and decision consolidation #1: File extraction to .ai-team/lib/ — 6h
- Recruit beta cohort (5-10 users via insider channel)
- Blog post feat: GitHub Issues intake, PRD mode, and human team members #2 (Feb 27-28): Under the hood — single .squad/ folder, lib/ split
Week 3 (Mar 3-9): Implementation Wave 2 + Beta Testing
- Rename .ai-team/ to .squad/ with backward-compatible migration #69 PR feat: GitHub Issues intake, PRD mode, and human team members #2: Documentation mass update (~120 files) — 5h
- Rename .ai-team/ to .squad/ with backward-compatible migration #69 PR [TEST] API capability test — will be deleted #3: Workflows dual-path detection — 6h
- Squads in GitHub Enterprise Copilot- Invalid config "Prompt" exceeds max length 30000 #76 PR feat: GitHub Issues intake, PRD mode, and human team members #2: Validation testing — 6h
- Beta testing with 7 exit criteria validation
- Bug fixing including Squad undid its own uncommited changes from a previous session #86 validation
- Continuous insider builds
- Blog post [TEST] API capability test — will be deleted #3 (Mar 6-7): Insider testing signals
Week 4 (Mar 10-16): Final Wave + Release
- Remove .ai-team/ deprecation banner (no longer needed after rename) #71, Standardize timestamps to ISO 8601 UTC across all logs and decisions #84, Release & CI/CD Hardening Plan for v0.5.0 #62 completion (cleanup work)
- Final validation (all 7 criteria + Squad undid its own uncommited changes from a previous session #86)
- Beta graduation
- Release v0.5.0 (March 16)
- Blog post Feature: Issue Assignment & Auto-Triage for Squad Members #4 (March 16): v0.5.0 ships
Success Criteria
- All 6 MUST ship issues resolved
- All 7 beta exit criteria met
- Squad undid its own uncommited changes from a previous session #86 fix validated (no regression)
- Migration tested on 5+ real repos
- Blog posts completed (4 total)
- Insider Program launched Week 1
- Documentation complete (migration guide, insider setup)
Week 1 Day 2 Progress (Feb 18)
Completed:
- ✅ Squad undid its own uncommited changes from a previous session #86 investigation (Fenster: reproduced, prompt-only fix, 3-4h estimate)
- ✅ Rename .ai-team/ to .squad/ with backward-compatible migration #69 audit (Fenster: 1,672 refs → 3 atomic PRs over 2 weeks)
- ✅ Squads in GitHub Enterprise Copilot- Invalid config "Prompt" exceeds max length 30000 #76 architecture (Verbal: 108KB → lib/ split, 12-16h estimate)
- ✅ PR feat: add 'squad watch' — Ralph local watchdog for persistent polling #57 review (Keaton: community contribution, defer to Week 2)
In Progress:
- 🔄 Insider Program setup (PRIMARY Week 1 task — NOT STARTED YET)
- 🔄 Blog post feat: worktree awareness, Scribe auto-commit, and decision consolidation #1 drafting (McManus — DUE FEB 20-21, 2 days away)
Next:
- Route Insider Program setup to Kobayashi (CI/CD + branch automation)
- Route blog post feat: worktree awareness, Scribe auto-commit, and decision consolidation #1 to McManus (insider program launch announcement)
Communications
- Insider Program: Binary model (insider or release), launches Week 1 for incremental testing
- Migration: Single
.squad/folder, mandatory but supported, backward compat until v1.0.0 - Squad undid its own uncommited changes from a previous session #86 fix: Trust signal — Squad protects your work (prompt-only, 3-4h, ships v0.5.0)
- Last breaking change: v1.0.0 will be polish, not breaking
Note: This is the last breaking change before v1.0.0. Post-v0.5.0 releases focus on stability, polish, and feature additions without migration pain.
Activity
bradygaster commentedon Feb 17, 2026
🎯 Insider Program & Release Cadence (PERMANENT INFRASTRUCTURE)
v0.5.0's mandatory beta program revealed the need for ongoing testing infrastructure with predictable release cadence. Three-agent consultation (Keaton: program structure, McManus: community engagement, Kobayashi: release automation) designed permanent systems.
Program Overview
Squad Insiders — 5-10 (growing to 20+) experienced users who test pre-releases 7-10 days before public ship, validate exit criteria, and give structured feedback.
Benefits:
#squad-insidersDiscord, fast-tracked issues)Commitment:
Entry paths:
Exit: Opt-out anytime (no penalty). Inactive for 3+ releases → alumni status (keep Discord, lose early builds, can rejoin).
v0.5.0 Integration
Beta testers become inaugural insiders. At v0.5.0 ship (March 16), invite each beta tester to join permanent program. Opt-in, same access/feedback loop, now ongoing.
Cohort growth: Start 5-10 (v0.5.0), add 3-5 per major release, cap at ~30.
Release Cadence
Pre-1.0 (Current):
Post-1.0 (Future):
Three-Tier Testing Process
Automation Roadmap (3 Phases)
Phase 1 (v0.5.0 — Build Now):
v0.5.0-beta.1,v0.5.0-beta.2)Phase 2 (v0.6.0 — Post-Beta Learnings):
Phase 3 (v1.0.0 — Polish for Production):
Community Engagement
Recruitment:
Recognition:
Communication:
insider-feedbacklabel: blockers, urgent feedbackOnboarding:
Implementation Timeline
Success Metrics
Recruitment: 10 by March 1, 20 by June 1, 50 by December 1
Engagement: 70%+ test each pre-release, 5+ actionable feedback per release, 50%+ quarterly retro participation
Recognition: 100% insiders credited in release notes, 3+ insider blog posts by EOY
Retention: <30% churn per quarter, alumni tier prevents hard drop-off
Open Questions for Brady
This infrastructure makes Squad's beta testing repeatable, scales from solo maintainer to distributed team, and creates a contributor pipeline: insiders → contributors → maintainers.
Full Recommendations:
.ai-team/decisions/inbox/keaton-insider-program-structure.md.ai-team/decisions/inbox/mcmanus-insider-community-engagement.md.ai-team/decisions/inbox/kobayashi-release-cadence-testing.mdbradygaster commentedon Feb 17, 2026
📦 Shipped Enhancement — PR #80
docs: add llms.txt outputs to docs build (merged Feb 17 by @danielscholl)
Adds LLM-discoverable documentation outputs:
llms.txt— curated docs index (pages LLMs should prioritize)llms-full.txt— full corpus export (all Squad docs in one file)*.html.md— markdown companions aligned to rendered page URLsWhy it ships with v0.5.0: Enhances discoverability for AI tooling that wants to learn Squad. Low-risk documentation enhancement, already merged and tested.
Implementation: Updated
docs/build.jswith shared frontmatter stripping helper, clean snippet extraction for readable descriptions.bradygaster commentedon Feb 17, 2026
📦 Shipped Enhancement — PR #89
fix: mcp discord notifications setup instructions (merged Feb 17 by @digitaldrummerj)
Fixes Issue #88 (Discord docs outdated) - updates MCP Discord notification setup instructions.
Impact on v0.5.0 scope: Removes #88 from deferred list - already fixed by community contribution. One less item to carry into v0.6.0.
bradygaster commentedon Feb 17, 2026
🔍 Architectural Analysis: Issues #86 and #87
✅ Issue #86: Data Loss Bug (CONFIRMED HIGH SEVERITY — Stay in Scope)
Root cause: Prompt engineering failure — agents lack git discipline. When hitting errors, agents use
git checkoutto undo without checking for other uncommitted changes → wipes out prior work from same session.NOT migration-related. This bug exists in v0.4.1 right now. Directory rename has zero impact on git operations.
Action plan:
❌ Issue #87: Workflows Assume npm (Stay Deferred to v0.6.0)
Root cause: Template generation doesn't detect project type. Squad was built FOR Squad (npm), so templates reflect that. Non-npm projects (Python, .NET, Java, Go) hit wrong assumptions.
Architectural verdict: NOT architectural. Doesn't affect Squad's core (agents, memory, coordinator) — it's template polish.
Why defer:
If pulling into v0.5.0: Add 2 weeks to timeline (Mar 16 → Mar 30), recruit multi-language beta, accept diluted feedback. Not recommended.
Summary
v0.5.0 scope unchanged. Existing plan is correct.
bradygaster commentedon Feb 20, 2026
v0.5.0 Sprint Complete ✅
All work items shipped:
Final test count: 64/64 passing on dev.
v0.5.0 ready for release merge (dev → preview → main).