feat: worktree awareness, Scribe auto-commit, and decision consolidation by amolchanov · Pull Request #1 · bradygaster/squad

amolchanov · 2026-02-07T18:32:34Z

Summary

Adds worktree support and Scribe improvements so that Squad works correctly in multi-worktree setups.

Changes

Worktree Awareness (squad.agent.md)

Two strategies: worktree-local (branch-isolated state, recommended for concurrent work) and main-checkout (shared state, single-session only)
Auto-detection: checks if .ai-team/\ exists in current worktree; falls back to main checkout if not
Coordinator resolves team root once and passes \TEAM_ROOT\ into every spawn prompt — agents never discover it themselves

Scribe Auto-Commit (squad.agent.md, scribe-charter.md)

Scribe commits .ai-team/\ changes after every session with a detailed \docs(ai-team):\ conventional commit message
Itemizes: session logged, decisions merged, cross-agent updates propagated

Decision Consolidation (squad.agent.md, scribe-charter.md)

After merging inbox, Scribe deduplicates decisions.md:
- Exact duplicates (same heading) → keep first
- Overlapping decisions (same topic, different authors/dates) → consolidate into a single (consolidated)\ block with merged rationale

Merge-safe append-only files (squad.agent.md, index.js)

.gitattributes\ \merge=union\ rules for decisions.md, history.md, log/, orchestration-log/
\index.js\ auto-creates these rules during
px squad\ install

Template updates (charter.md, scribe-charter.md)

Both templates include worktree awareness guidance and reference \TEAM_ROOT\

Files changed

.github/agents/squad.agent.md\ — main agent instructions
\index.js\ — install script (.gitattributes setup)
\ emplates/charter.md\ — agent charter template
\ emplates/scribe-charter.md\ — Scribe charter template

- Add Worktree Awareness section with two strategies (worktree-local, main-checkout) - Coordinator resolves team root and passes TEAM_ROOT to all spawn prompts - Scribe commits .ai-team/ changes with detailed conventional commit messages - Scribe deduplicates and consolidates overlapping decisions after merge - Add .gitattributes merge=union rules for append-only .ai-team/ files (docs + index.js) - Update charter.md and scribe-charter.md templates with worktree and commit guidance

- Add Ceremonies section to squad.agent.md with facilitator pattern - Ceremonies are configurable per squad via .ai-team/ceremonies.json - Support auto and manual triggers (before/after work batches) - Facilitator spawns participants as sub-tasks, synthesizes summary - Two default ceremonies: Design Review (before) and Retrospective (after) - Add ceremonies.json template and auto-copy in index.js - Add ceremony signals to routing table - Add ceremonies.json to Source of Truth hierarchy

- Instruct facilitator to keep ceremonies fast and focused - Participants provide domain-specific concerns, interface needs, and blockers only - Goal: minimize iterations by surfacing problems before independent work begins - Explicitly prohibit rehashing known context — ask for delta feedback only

Root cause: git -C with Windows paths and multiline git commit -m with backtick-n both fail silently in PowerShell. Fix: - Use cd into team root instead of git -C - Write commit message to temp file, commit with git commit -F - Add verification step: git log --oneline -1 after commit

After a ceremony completes, skip auto-triggered ceremony checks for the next step. Prevents back-to-back ceremonies (e.g., before-ceremony completing then immediately triggering an after-ceremony check). Cooldown resets after one batch of work completes without a ceremony.

Consistent with rest of .ai-team/ which is all markdown. No programmatic consumer needs structured JSON — the Coordinator parses ceremonies.md as natural language like every other config file.

bradygaster · 2026-02-07T22:30:17Z

thank you for the pull request @amolchanov! you're the first! i do have one question - and this is more a behavioral nitpick - I am currently working with Squad on two projects; one of them i've explicitly .gitignored the Squad-related stuff for reasons but, the intent of the product is, obviously, for the team to be a part of the repo. So, in spirit i agree with you but i can imagine there are some scenarios in which a user might want to keep their code separate from their Squad and do the same sort of thing I've done. In that case, would your tweak be a no-op or would it try to "un-ignore" the files? If you could do a few tests of that scenario, I do think that would be a valid concern for some users. I'd suggest making Scribe (I guess it was a matter of time before someone referenced Scribe in first person) decide intelligently on that commit behavior if the user has explicitly added the team files to the ignore list. Scribe should probably just see that and presume that's a step they can avoid.

Aside from that, I'm eager to take the contribution - just curious if you'd mind testing that one odd edge case. Should Scribe be aggressive and force-unignore the files? I'd want us to be a little careful. At the same time, if users - or you - think I should lean in hard that "the squad is a part of your repo and grows with it," I could take it as constructive product owner feedback. ;)

bradygaster · 2026-02-07T22:52:45Z


+### Worktree Awareness
+
+Squad and all spawned agents may be running inside a **git worktree** rather than the main checkout. All `.ai-team/` paths (charters, history, decisions, logs) MUST be resolved relative to a known **team root**, never assumed from CWD.


This is hot. Thank you! I'd been wondering how to bring worktrees to Squad. This is so great.

bradygaster · 2026-02-07T22:57:34Z

Never mind my concern - I see why this is an all-or-nothing and it is absolutely the right direction. Merged!

amolchanov · 2026-02-08T16:28:23Z

Hi Brady, I actually have a bug in my PR, the intent was to have scribe commit everything: all modified code and squad files. This is the direction that most agentic frameworks are moving so that the user can easily see changes in git and undo them for each iteration. I was just thinking that we might need to be even more aggressive and do multiple commits. I.e. create a worktree per turn, let all squad members commit their changes as they are done so that you can see exactly who was doing what. Maybe even working tree per squad member as it would be in the real life. I was experimenting yesterday with squad trying to build a simple Unity game and things went downhill fast. This is mostly due to complexity of Unity and every turn was breaking as many things as it was fixing, so it is absolutely a requirement to be able to see changes for each turn. I don't have experience with Unity, so that was probably a factor too. The log the scribe was keeping could have been contributing to it because I didn't check what it was writing to it and it could have been causing the model to go haywire and keep reintroducing the same issue. I need to confirm this and delete those files to see if it fixes the issue that it introduced again. If you want to keep this idea we need to figure out how ground those files to be accurate (i.e. things that worked or not). Happy to chat on Teams tomorrow if you would like to discuss this further. Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: Brady Gaster ***@***.***> Sent: Saturday, February 7, 2026 2:30:43 PM To: bradygaster/squad ***@***.***> Cc: amolchanov ***@***.***>; Mention ***@***.***> Subject: Re: [bradygaster/squad] feat: worktree awareness, Scribe auto-commit, and decision consolidation (PR #1) [https://avatars.githubusercontent.com/u/41929050?s=20&v=4]bradygaster left a comment (bradygaster/squad#1)<#1 (comment)> thank you for the pull request @amolchanov<https://github.com/amolchanov>! you're the first! i do have one question - and this is more a behavioral nitpick - I am currently working with Squad on two projects; one of them i've explicitly .gitignored the Squad-related stuff for reasons but, the intent of the product is, obviously, for the team to be a part of the repo. So, in spirit i agree with you but i can imagine there are some scenarios in which a user might want to keep their code separate from their Squad and do the same sort of thing I've done. In that case, would your tweak be a no-op or would it try to "un-ignore" the files? If you could do a few tests of that scenario, I do think that would be a valid concern for some users. I'd suggest making Scribe (I guess it was a matter of time before someone referenced Scribe in first person) decide intelligently on that commit behavior if the user has explicitly added the team files to the ignore list. Scribe should probably just see that and presume that's a step they can avoid. Aside from that, I'm eager to take the contribution - just curious if you'd mind testing that one odd edge case. Should Scribe be aggressive and force-unignore the files? I'd want us to be a little careful. At the same time, if users - or you - think I should lean in hard that "the team is a part of your repo and grows with it," I could take it as constructive product owner feedback. ;) — Reply to this email directly, view it on GitHub<#1 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAEMPUUKCDZJKXITSWYTA6D4KZRQ7AVCNFSM6AAAAACUKZ74S6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQNRVGY2TQMZQGU>. You are receiving this because you were mentioned.

* test: TDD contract tests for autonomous-pipeline core split Add 31 tests defining the behavioral contract for the planned autonomous-pipeline-core module (PRD Recommendation #1). Tests cover: - Task→tier assignment via selectResponseTier with DEMO_CONFIG - CostTracker per-agent accumulation across multiple agents/tasks - Agent state transitions (idle→working→idle→done lifecycle) - selectResponseTier routing properties (tier, modelTier, maxAgents, timeout) - Task queue creation, role coverage, and findNextTask assignment logic - simulateCost model contract (complexity→tokens→cost scaling) - Core pipeline wiring without telemetry/dashboard/OTel dependencies All 31 tests pass against current code. After the refactor splits core logic from telemetry+dashboard, these same assertions verify the extracted module preserves identical behavior. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * samples: standardize READMEs, naming, and add template - Standardize all 9 package.json names to @squad-samples/* format (Rec #3) - Rewrite root samples/README.md as public-facing index with learning path (Rec #9) - Standardize all 9 per-sample READMEs to consistent template (Rec #8) - Apply Microsoft Writing Style Guide across all sample documentation - Add SAMPLE-README-TEMPLATE.md for future sample authors - Add 31 TDD tests for autonomous-pipeline core split (Rec #1 prep) 17 files modified, 1 new file. All markdownlint and cspell checks pass. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Merged 9 inbox decisions → decisions.md (personal squad impl plan, economy mode design/governance, persistent model preference, user directives, template sync, ESM fix) - Deleted inbox files (merge complete) - Updated agent histories: Flight (personal squad plan), EECOM (economy mode), Procedures (governance proposals) - Appended team updates to agent contexts Fleet Status: - Flight: Personal squad design validated, 19-task plan across 4 PRs (MVP = PR #1+3) - EECOM: Economy mode complete (34 tests), PR #504 open - Procedures: Economy + consult-mode governance drafted, PR #503 open - Directives: #464 rate limit UX (soft dep #500), #502 bug P1 next-up, triage includes discussions Ready for execution: EECOM Phase 1, Procedures Phase 3 (concurrent). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat: economy mode skill, governance proposals for #500 and #344 - Add .squad/skills/economy-mode/SKILL.md: Layer 3 cost modifier - Economy table shifts code tasks to gpt-4.1/gpt-5-mini, arch to sonnet - Activation: session phrase, config.json economyMode:true, or --economy flag - Layer 0/1/2 (user intent) always overrides economy mode - 💰 indicator in spawn acknowledgments - Confidence: low (first implementation) - Add .squad/decisions/inbox/procedures-economy-mode-governance.md - Draft additions to squad.agent.md for Flight review - Economy mode paragraph after Layer 0, updated table, 💰 convention - Valid models catalog audit findings (claude-sonnet-4.6 missing, etc.) - Add .squad/decisions/inbox/procedures-personal-squad-governance.md - Five gap analysis: Init Mode, consult mode detection, TEAM_ROOT semantics, charter templates, missing consult-mode skill - Proposed CONSULT_MODE spawn signal and 🧳 acknowledgment format - Proposed new consult-mode skill (post-governance-approval) - Update procedures history.md with learnings Closes #500 References #344 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat: economy mode for cost-conscious model selection (#500) Add economy mode that shifts model selection to cheaper alternatives at Layer 3 (task-aware auto) and Layer 4 (default). Explicit user preferences at Layers 0-2 are never overridden. SDK changes: - Add ECONOMY_MODEL_MAP and applyEconomyMode() to config/models.ts - Add readEconomyMode() / writeEconomyMode() for config.json persistence - Update resolveModel() in config/models.ts with economyMode option - Update resolveModel() in agents/model-selector.ts with economyMode option CLI changes: - Add 'squad economy [on|off]' command for persistent toggle - Add --economy global flag for session-level activation (SQUAD_ECONOMY_MODE=1) Tests: - 34 new tests in test/economy-mode.test.ts (all passing) Economy model table: claude-opus-4.6 -> claude-sonnet-4.5 (architecture/review) claude-sonnet-4.6 -> gpt-4.1 (code writing) claude-haiku-4.5 -> gpt-4.1 (docs/mechanical) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Scribe: Wave 1 decisions merged (#329/#344/#500) - Merged 9 inbox decisions → decisions.md (personal squad impl plan, economy mode design/governance, persistent model preference, user directives, template sync, ESM fix) - Deleted inbox files (merge complete) - Updated agent histories: Flight (personal squad plan), EECOM (economy mode), Procedures (governance proposals) - Appended team updates to agent contexts Fleet Status: - Flight: Personal squad design validated, 19-task plan across 4 PRs (MVP = PR #1+3) - EECOM: Economy mode complete (34 tests), PR #504 open - Procedures: Economy + consult-mode governance drafted, PR #503 open - Directives: #464 rate limit UX (soft dep #500), #502 bug P1 next-up, triage includes discussions Ready for execution: EECOM Phase 1, Procedures Phase 3 (concurrent). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: hard-fail on Node <22.5.0 with clear message (#502) node:sqlite (required by the Copilot SDK for session storage) was added in Node 22.5.0. Previously Squad only warned and continued, leading to cryptic ERR_UNKNOWN_BUILTIN_MODULE crashes later. Changes: - cli-entry.ts: replace soft warn+continue with hard process.exit(1) and a clear upgrade message when Node <22.5.0 is detected at startup. Remove now-redundant checkNodeSqlite() function and its call site. - doctor.ts: add checkNodeVersion() check to squad doctor diagnostics so users can detect the version gap proactively. - package.json (root, squad-cli, squad-sdk): correct engines field from >=20 to >=22.5.0, enabling npm/npx to warn at install time. - test/cli/doctor.test.ts: add 5 tests for checkNodeVersion pass/fail paths; update check-count assertion. Closes #502 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Add 'squad personal init/list/add/remove' subcommands - Add 'squad cast' to show merged session cast - Add --team-root CLI flag for explicit team root override - Register commands in cli-entry.ts with help text PR #2 of #508 implementation plan. Depends on: squad/508-sdk-foundation (PR #1) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat(sdk): add personal squad resolution and ambient discovery (#508) - Add personalDir to ResolvedSquadPaths - Add resolvePersonalSquadDir() with SQUAD_NO_PERSONAL kill switch - Add PersonalAgentMeta/PersonalAgentManifest types - Add resolvePersonalAgents() for personal agent discovery - Add mergeSessionCast() to merge personal into project cast - Add ensureSquadPathTriple() for triple-root path validation - Export all new functions and types Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat(cli): add personal squad commands and cast display (#508) - Add 'squad personal init/list/add/remove' subcommands - Add 'squad cast' to show merged session cast - Add --team-root CLI flag for explicit team root override - Register commands in cli-entry.ts with help text PR #2 of #508 implementation plan. Depends on: squad/508-sdk-foundation (PR #1) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Add preflight job scanning all packages for file: references and validating semver versions before any publish work begins - Add npm pack --dry-run validation step to smoke-test job - Disable ghost publish-npm.yml workflow via GitHub API - Add purpose documentation to ci-rerun.yml - Fix YAML parse issue: quote step names containing 'file:' colon - Pipeline chain: preflight → smoke-test → publish-sdk → publish-cli This is the #1 gate that would have prevented the v0.9.1 incident. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Write orchestration log: 2026-03-22T17-30-gandalf.md (Teams watchdog decomposition outcome) - Write session log: 2026-03-22T17-30-teams-watchdog-decomp.md (brief session summary) - Merge 3 decision inbox files into decisions.md (copilot CLI directive, scripting philosophy, watchdog architecture) - Delete inbox files after merge - Update gandalf/history.md with Teams watchdog learnings Issues bradygaster#1-bradygaster#6 created on jbenami_microsoft/ms-pa for 6-step watchdog pipeline. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Elrond: Skills scan of tamirdresher/squad-skills (20 plugins) * Identified 3 directly relevant: teams-monitor, news-broadcasting, secrets-management * Documented full inventory with P0/P1 prioritization * Key insight: WorkIQ poll-based, well-suited for daily batch design - Bilbo: Comprehensive plugin catalog * Created docs/squad-skills-catalog.md (25.2 KB, 21 plugins) * Mapped to Teams watchdog Issues bradygaster#1-bradygaster#6 * Organized by skill domain, implementation order - Scribe: Decision management * Merged .squad/decisions/inbox → decisions.md * Wrote orchestration logs for Elrond, Bilbo * Wrote session log for skills scan Teams watchdog design constraints confirmed: rate-limit WorkIQ to 1 query/agent cycle. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Adds probe-messages.ps1 that fetches recent Teams messages using copilot -p with the workiq-ask_work_iq MCP tool. Configurable time window via -Hours parameter. Output is plain text to stdout. Key finding: copilot -p requires --allow-tool='workiq' for non-interactive MCP tool consent. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…alidation, test gaps Close all four FIDO blocking gaps and fold Flight + RETRO nits: FIDO Gap bradygaster#1 — Consult setup-mode success path: Add test exercising the guarded consult setup path going GREEN from a registered consumer repo with a personal squad available. Asserts config.json written and .git/info/exclude updated. FIDO Gap bradygaster#2 — .gitignore non-mutation: Extend consult (x2) and link failure tests to assert .gitignore and .git/info/exclude are not written when the resolver guard fires. FIDO Gap bradygaster#3 — Deferred (see decision drop): URL-without-clone-destination, clone failure, host-verify are not present in the current implementation and have no prior behavior to preserve. Deferred to the piece that introduces those features. FIDO Gap bradygaster#4 — Dead resolved variable: - consult and link dispatch: remove dead named variable, use inline boolean guard (gate-only semantics; runners are self-contained). - assign dispatch: store guard result and thread it into RunAssignOpts via new resolved field so guard and runner share context. - Runner falls back to internal resolve for direct (non-CLI) callers. - Threading proven by asserting resolved.path appears in output. Flight nit — Guard harmonization: All three commands now use dispatch-level guards in cli-entry.ts. Inline boolean check for commands whose runners need no resolved data; threaded struct for assign where the runner uses resolved.callsign and resolved.path. resolveSquadV2 alias in cli-entry.ts unchanged (required to distinguish from legacy sdk.resolveSquad call pattern). RETRO nit — Path validation: Add path.isAbsolute() and ..-segment checks on opts.cwd and opts.home in runAssignToCopilot before any file-write path uses them. Validate --home at CLI dispatch boundary as well. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ns + cross-agent updates Sims successfully addressed all 4 blocking FIDO gaps + Flight/RETRO nits in piece 08b revision cycle: - Gap bradygaster#1: Consult setup-mode success test added - Gap bradygaster#2: .gitignore non-mutation assertions added - Gap bradygaster#3: URL/clone/host-verify deferred to future work (documented rationale) - Gap bradygaster#4: Dead resolved variable threaded to assign runner Test results: 31 GREEN (17 legacy-resolver-migration + 14 consult); Build CLEAN; Scrub gate baseline unchanged. Decisions merged: Sims 08b revision decision from inbox to canonical decisions.md. Cross-agent updates: - EECOM: Lockout cycle closed - FIDO: Gap closure confirmed - Flight: Guard location nits folded into revision - RETRO: All hardening candidates applied Logs written: - .squad/orchestration-log/2026-05-14T16-12-01-sims.md - .squad/log/2026-05-14T16-12-01-piece-08b-sims-revision.md Phase C ready: Squash pending. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Five fixes applied based on the tamresearch1 squad's content review (Picard, Q, Crusher, Troi, Guinan, Geordi, Paris, Neelix): 1. Slide 13 (classify spotlight) — corrected the classifier truth. The (CI|PR|build) regex sits in BOTH FORBIDDEN_PATTERNS AND the TRANSIENT heuristic; FORBIDDEN scan wins, so TRANSIENT branch is unreachable for those inputs (only fires via explicit --class). Removed incorrect "CI build always passes → TRANSIENT" example. Verified live against LocalMemoryStore.classify() on 2026-06-09. (Q's #1 ❌ FAIL finding.) 2. Slide 1 — changed "Co-creator of Squad" to "Contributor & maintainer" to avoid attribution risk noted by Q and Crusher (upstream creator is Brady Gaster per the tamresearch1 decisions.md). 3. Slide 16 — "ADC Squad runner demo" generalized to "External Squad-using project demo" (Crusher: unexplained acronym could read as internal codename in public talk). 4. Slide 4 — "dozens of changesets" expanded to "97 changesets across CLI and SDK (50 SDK + 71 CLI, deduplicated)" per CHANGELOG.md line 9 (Q: underquoting invites skepticism). 5. NEW slide 62 (page 61) — "The three questions, answered in human terms" closing synthesis that answers slide 02's opening questions in plain language. Bottom line: "your team doesn't lose context to keep the repo clean — it gets both." Addresses Troi/Guinan/Picard's structural feedback: the deck showed the mechanism honestly but was missing the relief. Plus: workflow stage 4 slide updated to note the FORBIDDEN/TRANSIENT regex overlap explicitly in the code comment block. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

github-actions · 2026-06-12T17:55:45Z

🛫 PR Readiness Check

ℹ️ This comment updates on each push. Last checked: commit 82109ba

PR Scope: 🔧 Infrastructure

⚠️ 3 item(s) to address before review

Status	Check	Details
❌	Single commit	6 commits — consider squashing before review
✅	Not in draft	Ready for review
✅	Branch up to date	Up to date with dev
❌	Copilot review	No Copilot review yet — it may still be processing
✅	Changeset present	No source files changed — changeset not required
✅	Scope clean	No .squad/ or docs/proposals/ files
✅	No merge conflicts	Merge status unknown — GitHub is still computing
✅	Copilot threads resolved	No Copilot review threads
❌	CI passing	1 check(s) still running

Files Changed (5 files, +365 −5)

File	+/−
`.github/agents/squad.agent.md`	+246 −3
`index.js`	+30 −0
`templates/ceremonies.md`	+41 −0
`templates/charter.md`	+2 −0
`templates/scribe-charter.md`	+46 −2

Total: +365 −5

This check runs automatically on every push. Fix any ❌ items and push again.
See CONTRIBUTING.md and PR Requirements for details.

…oard (#1373) * docs: add SWE-bench Lite technical report (198/300 = 66.0%) * fix: reconcile timeout numbers, add Limitations section, link public data repo Addresses Brady's review: - Clarify 31 timeouts total (20 no-patch + 11 partial-patch errors) - Remove wording inconsistency between Key Parameters and Error Analysis - Add Limitations & Attribution section (model vs orchestration) - Link to public tamirdresher/squad-swe-bench for full reproduction Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Brady Gaster <41929050+bradygaster@users.noreply.github.com>

amolchanov added 6 commits February 7, 2026 10:32

refactor: convert ceremonies from JSON to markdown

5b00169

Consistent with rest of .ai-team/ which is all markdown. No programmatic consumer needs structured JSON — the Coordinator parses ceremonies.md as natural language like every other config file.

bradygaster reviewed Feb 7, 2026

View reviewed changes

bradygaster merged commit 9b62931 into bradygaster:main Feb 7, 2026

This was referenced Feb 18, 2026

[EPIC] v0.5.0 — .squad/ Migration & Architectural Cleanup #91

Closed

Release & CI/CD Hardening Plan for v0.5.0 #62

Closed

Rename .ai-team/ to .squad/ with backward-compatible migration #69

Closed

diberry mentioned this pull request Mar 8, 2026

docs: add validation steps to Quick Start README (#285) #286

Merged

This was referenced Mar 20, 2026

Bug: history-shadow.ts has read-modify-write race condition (data loss) #479

Closed

PRD: Squad State — Typed StorageProvider Interface with Contract Conformance #481

Closed

samples: standardize READMEs, naming, and add template #484

Merged

bradygaster mentioned this pull request Mar 22, 2026

Implementation: Ambient Personal Squad (#329 + #344) #508

Closed

This was referenced Mar 26, 2026

fix: getPersonalSquadRoot() returns correct personal-squad path (#590) #620

Merged

discussion: Post-0.9.1 routing regression — community feedback #612

Closed

diberry mentioned this pull request Apr 7, 2026

PRD: Squad SDK Hardening — Architecture & Security Improvements #895

Closed

16 tasks

danielmeppiel mentioned this pull request Apr 26, 2026

APM integration — proposed target architecture (follow-up to #1015) #1049

Closed

joniba mentioned this pull request May 17, 2026

Bug: Personal squad fails to auto-detect #984

Closed

This was referenced May 19, 2026

Implement squad:// URI resolver #1137

Open

Add squad asset add — install a single atom from a remote registry #1141

Open

tamirdresher mentioned this pull request May 20, 2026

Squad telemetry from embedded host processes (Agent Framework wrappers, no REPL): how should it reach Aspire? #1144

Open

wbreza mentioned this pull request May 21, 2026

Add vally eval suite (copilot-sdk executor, real agent evals) #1115

Closed

tamirdresher mentioned this pull request May 27, 2026

Refresh squad.agent.md model catalog + dynamic catalog discovery #1183

Closed

tamirdresher mentioned this pull request Jun 5, 2026

Address pre-existing Copilot review comments surfaced by PR #1212 (workflow permissions, logic bugs, YAML escaping) #1213

Closed

bradygaster mentioned this pull request Jun 10, 2026

fix: address pre-existing Copilot review comments from PR #1212 #1258

Merged

tamirdresher mentioned this pull request Jun 11, 2026

arch: pick a direction for agents/{name}/history.md vs the governed memory model #1243

Open

This was referenced Jun 13, 2026

Coordinator does not probe for squad_state/memory tools — silently uses raw file ops on non-local state backends #1305

Closed

feat: add 'squad preset install <source>' for sharing presets via repo URL (#1224) #1225

Merged

tamirdresher mentioned this pull request Jun 23, 2026

docs: SWE-bench Lite technical report — 198/300 (66.0%) #1 on leaderboard #1373

Merged

4 tasks

bradygaster mentioned this pull request Jun 23, 2026

docs: add SWE-bench Lite results blog post (#032) #1374

Closed

bradygaster mentioned this pull request Aug 25, 2026

Allow dependency additions without bypassing protected-file safety #1748

Closed

Copilot AI mentioned this pull request Sep 2, 2026

feat(nap): measure charter/skill context and add --dry-run --json #1987

Closed

Navigation Menu

feat: worktree awareness, Scribe auto-commit, and decision consolidation - #1

feat: worktree awareness, Scribe auto-commit, and decision consolidation#1
bradygaster merged 6 commits into
bradygaster:mainfrom
amolchanov:worktree-awareness-and-scribe-commit

amolchanov commented Feb 7, 2026

Uh oh!

bradygaster commented Feb 7, 2026 •

edited

Loading

Uh oh!

bradygaster Feb 7, 2026

Uh oh!

bradygaster commented Feb 7, 2026

Uh oh!

amolchanov commented Feb 8, 2026 via email

Uh oh!

github-actions Bot commented Jun 12, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants


		### Worktree Awareness

		Squad and all spawned agents may be running inside a git worktree rather than the main checkout. All `.ai-team/` paths (charters, history, decisions, logs) MUST be resolved relative to a known team root, never assumed from CWD.

Conversation

amolchanov commented Feb 7, 2026

Summary

Changes

Files changed

Uh oh!

bradygaster commented Feb 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bradygaster Feb 7, 2026

Choose a reason for hiding this comment

Uh oh!

bradygaster commented Feb 7, 2026

Uh oh!

amolchanov commented Feb 8, 2026 via email

Uh oh!

github-actions Bot commented Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🛫 PR Readiness Check

⚠️ 3 item(s) to address before review

Files Changed (5 files, +365 −5)

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

bradygaster commented Feb 7, 2026 •

edited

Loading

github-actions Bot commented Jun 12, 2026 •

edited

Loading