Repository navigation
Description
Everything left before Claude Code support in Squad is done, from epic #2438. Phases 0, 2, 3, 4, 5 and 6 are merged (#2445, #2450, #2463). Phase 1 is merged in #2451 except for the live golden trace.
Each item is marked Must (blocks calling Claude support done for 2.0), Should (strongly wanted for 2.0) or Later (can ship after 2.0).
1. Finish Phase 1: execution harness (#2441)
- Must Merge test(backend): run the golden trace live on the claude-agent-sdk harness (#2441) #2452: the G1–G9 golden trace, run live on
claude-agent-sdk(merged 2026-11-12) - Must Close Claude host PR 2: Claude Agent SDK HarnessAdapter (#2438 Phase 1) #2441 (closed)
- Later (deferred 2026-11-13) Maintainers decide: pending approvals live in memory, and a harness restart cancels them and resumes the session. Does that meet "durable approval versions" (harness-adapters §15 item 6)? If not, persist pending approvals
- Should "Make
@github/copilot-sdkan optional peer dependency" (epic Phase 1): filed as Make @github/copilot-sdk an optional peer dependency of squad-sdk (#2438) #2472
2. Bugs from the end-to-end run
- Must fixed in fix(claude): portable launch commands and host-aware init next steps (#2462, #2453) #2470 (merged): squad init commits machine-specific node paths into .mcp.json and .claude/settings.json #2462
squad initcommits machine-specificnode <abs path>launchers into.mcp.jsonand.claude/settings.jsonhooks, which breaks the repo for anyone else who clones it - Must fixed in fix(claude): portable launch commands and host-aware init next steps (#2462, #2453) #2470 (merged): squad init --client claude tells Claude users to run copilot --agent squad #2453
squad init --client claudetells Claude users to runcopilot --agent squad - Should fixed in fix(claude): refresh members after routing, clearer health dirs, no retired logs, plugin validates clean (#2456, #2455, #2454, #2458) #2471 (merged): Claude member subagent descriptions are written before routing exists and stay vague #2456 Member subagent descriptions are written at cast time, before routing exists, and stay vague until
squad upgrade - Should fixed in fix(claude): refresh members after routing, clearer health dirs, no retired logs, plugin validates clean (#2456, #2455, #2454, #2458) #2471 (merged): Coordinator and Scribe's charter disagree on who writes session and orchestration logs #2454 The coordinator asks Scribe for session and orchestration logs, but Scribe's charter rules them out
- Should fixed in fix(claude): refresh members after routing, clearer health dirs, no retired logs, plugin validates clean (#2456, #2455, #2454, #2458) #2471 (merged): squad_state_health resolvedTeamDir reports the repo root, not the squad directory #2455
squad_state_healthresolvedTeamDirreports the repo root, not.squad - Should fixed in fix(claude): refresh members after routing, clearer health dirs, no retired logs, plugin validates clean (#2456, #2455, #2454, #2458) #2471 (merged): squad init --client claude also writes .copilot, .github and .vscode: intended? #2458
squad init --client claudealso writes.copilot/,.github/and.vscode/. Confirm it's intended and document it - Later Squad stamps dates in UTC, so team files and logs show tomorrow's date in US evenings #2457 Dates are stamped in UTC (a day ahead in US evenings). Not Claude-specific
3. Plugin and dashboard hardening (after #2463)
- Must regenerated against
2.0.0-preview.5in release: prepare 2.0.0-preview.5 #2506 (release prep): The plugin and marketplace pin2.0.0-preview.4, which lacks the dashboard and hook commands. They must be regenerated against the CLI version that ships these features (open thread on feat(claude): Squad Claude Code plugin and team dashboard (#2443, #2459, #2460) #2463). This happens as part of the release, see §6 - Must Install the plugin for real from the GitHub marketplace on a clean machine; so far only
--plugin-dirhas been run. Can run once 2.0.0-preview.5 is published - Should fixed in fix(claude): repository launcher finds squad on PATH; launcher argv tested on Windows #2483 (merged; Windows CI runs the launcher tests): Test the Windows launcher's
npxquoting. Also fixes the repository launcher never findingsquadon PATH (a fix(claude): portable launch commands and host-aware init next steps (#2462, #2453) #2470 regression) - Should fixed in fix(claude): refresh members after routing, clearer health dirs, no retired logs, plugin validates clean (#2456, #2455, #2454, #2458) #2471 (merged): Add frontmatter to the
cli-wiringandretro-enforcementtemplate skills (claude plugin validatewarnings) - Later Plugin icon
- Later Team dashboard as a canvas in the GitHub Copilot app #2461 Team dashboard as a canvas in the GitHub Copilot app
4. Phase 7: host-neutral canonical coordinator
- Should decided: option C, ADR-003 Accepted in docs(adr-003): accept option C #2501; Claude 2.0: self-sufficient non-Copilot coordinators (ADR-003 option C) #2487 done (live scenarios with the Copilot file absent recorded in the host feature matrix, docs(matrix): record the #2487 live scenarios on Claude Code #2503); option B stays an opt-in spike in Spike: host-neutral canonical coordinator (ADR-003 option B), opt-in experiment #2488. ADR-003 direction from the team (2026-11-12): option C for 2.0 → Claude 2.0: self-sufficient non-Copilot coordinators (ADR-003 option C) #2487 (self-sufficient non-Copilot coordinators, live behavioral scenarios with the Copilot file absent, about 3–5 days); option B as an opt-in spike → Spike: host-neutral canonical coordinator (ADR-003 option B), opt-in experiment #2488 (fixed budget, not shipped). Analysis: docs(adr-003): gains, risks and scope for the host-neutral coordinator #2486. Original item: Decision (not a release gate) Analysis in docs(adr-003): gains, risks and scope for the host-neutral coordinator #2486 recommends option C (drop the pointer to the Copilot file, about half a day) and deferring B (about 2–3 days). Decide ADR-003 (
docs/specification/adr-003-host-neutral-canonical-coordinator.md, Proposed): does the canonical coordinator move from.github/agents/squad.agent.mdto a host-neutral source? Also decide whether it gates 2.0 or follows it - Implementation merged (Claude 2.0: self-sufficient non-Copilot coordinators (ADR-003 option C) #2487 work, fix(remote): a project linked to a team repository uses that team (satellite repos) #2500, docs(adr-003): accept option C #2501, docs(matrix): record the #2487 live scenarios on Claude Code #2503)
5. Review, test and docs loose ends
- Should done (threads resolved or annotated): Resolve or annotate the declined code-scanning threads on merged PRs: feat(claude): Claude projection v2: member subagents, skills, governance hooks, watch preset #2450 (hook launcher allowlist, hook-context text, member
Toolsallowlist) and feat(backend): claude-agent-sdk and copilot-sdk execution harnesses (#2441) #2451 (env allowlist for the Claude harness) - Should Run the epic's Phase 1 evidence item against a direct Anthropic API, Bedrock or Vertex endpoint (not a Copilot bridge), and record the endpoint in the evidence string
- Later A live approval test for the Copilot harness (only plain turns have run live)
- Later Replace the fake
claudebinary intest/cli/watch-claude-preset.test.tswith live runs behindSQUAD_REAL_CLI_TESTS=1 - Later Run a real gh-aw workflow with
engine: claudeandengine.args: ["--agent", "squad"] - Later decided and fixed in fix(claude): don't project Copilot's template Init Mode skills #2504 (merged): Claude no longer gets Copilot's template Init Mode skills. Decide whether to stop projecting the template
init-modeskill for Claude
Also found while working these
- The upgrade matrix's 1.0.1 and 1.1.0 release rows were silently skipping on
dev: the fixture fetched assets frombradygaster/squad, which now returns 404. Fixed in fix(claude): portable launch commands and host-aware init next steps (#2462, #2453) #2470, which points them atgithub/squadand refolds the matrix
6. Release gate (prepared in #2506, not yet dispatched)
- Must agreed: the next release is 2.0.0-preview.5, prepared in release: prepare 2.0.0-preview.5 #2506 (plugin and marketplace regenerated there). A maintainer merges release: prepare 2.0.0-preview.5 #2506 and dispatches
squad-release.yml. Agree the 2.0 release plan with Tamir: the version (another preview or 2.0), what ships in it, and who regenerates the plugin and marketplace against it. Nothing in this checklist starts the release process - Must After the release: tick the epic's success criteria and close Epic: Claude Code as a first-class Squad 2.0 host, and the non-Copilot host that promotes the v0.2 profiles #2438
🤖 Generated with Claude Code
2026-11-12: demo run findings
A real claude -p run on a fresh repository, recorded for the README:
- merged fix(claude): first-run fixes from the demo run: inline casting, member descriptions, dashboard end of session #2484 The coordinator skipped the init-mode skill and hand-wrote
.claude/agents/(blocked by Claude Code): Init Mode now namessquad cast/squad client syncinline. Casting went from 54 turns, $1.49 and no subagents to 17 turns, $0.37 and all subagents written - merged fix(claude): first-run fixes from the demo run: inline casting, member descriptions, dashboard end of session #2484 Member descriptions doubled a period, and the coordinator passed paragraphs as
--role - merged fix(claude): first-run fixes from the demo run: inline casting, member descriptions, dashboard end of session #2484
squad dashboardshowed a headless session as running forever (no Stop hooks → no end marker); it now ends atcost-state - merged docs(readme): See it in Claude Code, with a recorded team dashboard run #2485 README "See it in Claude Code" with the recorded dashboard GIF (trending goal)
2026-11-16: status
All code work for Claude Code support is merged. Merged 2026-11-13/14: #2496 (narrower Squad Bash allowlist), #2500 (linked team repositories), #2501 (ADR-003 option C accepted), #2502 (squad_decide says where a git-branch backend stored it), #2503 (live scenario evidence), #2504 (no Copilot Init Mode skills for Claude), #2505 (a user's team decision reaches decisions.md in the same turn, so a teammate on Copilot sees it; one team with one person on Claude Code and another on Copilot CLI checked live both ways).
What's left for this issue:
- Must (maintainer): merge release: prepare 2.0.0-preview.5 #2506 and dispatch the 2.0.0-preview.5 release
- Must: the clean-machine marketplace install, after the release
- Must: tick the epic's success criteria and close Epic: Claude Code as a first-class Squad 2.0 host, and the non-Copilot host that promotes the v0.2 profiles #2438
- Should: Phase 1 evidence against a direct Anthropic API, Bedrock or Vertex endpoint
Activity
bradygaster commentedon Nov 12, 2026
ADR-003 evidence: does the Claude coordinator read
.github/agents/squad.agent.md?ADR-003 option B rests on one cost: a non-Copilot coordinator may read the 80 KB Copilot coordinator on demand, because its digest says "if
.github/agents/squad.agent.mdexists in this workspace, read it as the authoritative coordinator contract". This comment checks whether that happens. The analysis is in #2486.Method. I went through every local Claude Code session transcript for a Squad test repository (
~/.claude/projects/<repo>/*.jsonl, Claude Code 2.1.317–2.1.318, 2026-11-10 to 2026-11-12) and checked every tool call for.github/agents/squad.agent.mdas a Readfile_path, a Bashcommand, or a Grep or Globpattern/path. The script is below.Result: 0 of 43 sessions and 0 of 223 tool calls touched the file.
Most of those sessions are short probes, so here are the ones that did real team work. In both repositories the 82–83 KB Copilot file was present, and the coordinator's digest pointed at it:
squad_state/.squad/accessessquad init --client claude, cast a team, build an MVP (22 tests pass)The coordinators worked from their digest, the
squad-coordinator-*skills andsquad_state. The same pattern shows in the demo transcripts: the coordinator readsteam.md,config.jsonandrouting.mdthroughsquad_state, then casts or dispatches.Caveats.
What it means for ADR-003. The pointer is close to dead text in practice, so B's stated gain is mostly theoretical. Option C, rewording the digest so it doesn't point at the Copilot file, matches how the coordinator already behaves.
Script
bradygaster commentedon Nov 13, 2026
ADR-003: team feedback and proposed direction
Reviewed the Teams discussion on 2026-11-12. All three substantive replies (Q, Bubu, and Davidbot) favor option C for 2.0, with option B explored separately as an isolated, unshipped spike. This matches Brady's earlier direction in the thread. To make the sequencing explicit: the replies recommend C now; consider shipping B later only if the experiment earns it, not delaying C until B is complete.
Related: analysis PR #2486; Claude Code epic #2438; Teams discussion.
C: deliver a complete Claude experience for 2.0
Make non-Copilot coordinator instructions self-sufficient without depending on
.github/agents/squad.agent.md. The user-visible goal is correct specialist delegation, preserved project policy and safeguards, and useful results through Claude-native tools—not merely a shorter prompt or a different file layout.Shared conditions from the replies:
The reported 43 sessions establish file-consumption observations, not behavioral completeness. Successful task completion alone is insufficient evidence that approval gates and backend rules were respected.
B: bounded experiment, not a shipping commitment
Prototype a host-neutral canonical workspace contract with Claude and Copilot projections in disposable, explicitly opted-in workspaces. No default installer migration or live-team migration as part of the spike. Give it a fixed experiment budget; C must proceed independently.
Reuse C's behavioral scenarios, then test fresh and customized legacy workspaces, monorepo/remote placement, repeatable upgrades, customization preservation and precedence, rollback that restores prior bytes, and no Copilot regression. Assess whether B measurably improves correctness or maintainability enough to justify migration risk—not just whether the directory layout is cleaner.
Outcome: pursue C for the 2.0 Claude experience; investigate B separately and return with evidence before any shipment decision. The replies are design recommendations, explicitly not independent implementation reviews. This comment records the discussion; it does not mark ADR-003 Accepted, approve an implementation, or authorize shipping B.