Skip to content

feat(skills): add SkillScriptLoader, handler types, and applySkillHandlers#325

Closed
EmmittJ wants to merge 3 commits intobradygaster:mainfrom
EmmittJ:EmmittJ/297-skill-script-loader
Closed

feat(skills): add SkillScriptLoader, handler types, and applySkillHandlers#325
EmmittJ wants to merge 3 commits intobradygaster:mainfrom
EmmittJ:EmmittJ/297-skill-script-loader

Conversation

@EmmittJ
Copy link
Copy Markdown
Contributor

@EmmittJ EmmittJ commented Mar 9, 2026 •
edited
Loading

Summary

Implements the executable skill handler layer described in PR #215 and referenced in #297. This adds the runtime loading and execution infrastructure that allows .squad/skills/{name}/ directories to contain JS handler scripts that replace built-in tool behavior at runtime.

Inspired by #297 (EmmittJ's design in PR #215)

What's included

New: handler-types.ts

Full type system for the skill-script model:

  • SkillHandler<T> — typed handler interface
  • HandlerLifecycle — init() / dispose() hooks
  • Concern interfaces: SearchConcern, ReadConcern, WriteConcern, ExecuteConcern
  • ConcernMap — maps handler concerns to tool names
  • LoadResult — what the loader returns after processing a skill directory
  • defineHandler() — typed helper for authoring handlers with full IDE support

New: skill-script-loader.ts

Runtime loader class:

  • SkillScriptLoader — scans scripts/ for .js files and derives tool names from filenames (create_issue.js → squad_create_issue). No hardcoded concern-to-tool mapping — the directory is the source of truth.
  • resolveSkillPath() — path containment check (guards against directory traversal)
  • wrapSkillHandler() — bridges SkillHandler → SquadToolHandler signature, injects backendConfig

Modified: ToolRegistry.applySkillHandlers()

Post-construction method that accepts SquadTool[] from a LoadResult and replaces matching registered tools with skill-provided handlers.

New: test/skill-script-loader.test.ts

33 tests covering: load, execute, error paths, missing scripts directory, invalid exports, lifecycle hooks.

Known limitations

  • LoadResult.tools is always [] for now — the loader discovers scripts and derives tool names correctly, but getToolSchema(toolName) returns undefined for all of them because those tool schemas (squad_create_issue etc.) don't yet exist in ToolRegistry. Once the tool schemas land, handlers wire up automatically with no loader changes needed.
  • ToolRegistry bootstrap wiring — callers must construct SkillScriptLoader with a getToolSchema callback and call applySkillHandlers() after ToolRegistry is initialized. Follow-on integration work needed to wire this into TrackingConfig → SquadConfig.

Testing

npm test

All pre-existing tests pass. 33 new tests added for the skill loader.

Co-authored-by: Copilot 223556219+Copilot@users.noreply.github.com

Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
fix: make sendAndWait timeout configurable (bradygaster#347)
c1d5c7c 
Closes bradygaster#325

Replaces hard-coded 120_000ms timeout with TIMEOUTS.SESSION_RESPONSE_MS (env: SQUAD_SESSION_TIMEOUT_MS, default 10 min).

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telem…
138652a 
…etry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
feat: rich progress indicators for agent work (bradygaster#357)
a7170d6 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 3 launch

Session: 2026-02-23T2228-phase3-launch
Phase 2 complete with P1 UX polish (PR bradygaster#356 merged).
Phase 3 Wave 1 launched: Cheritto (bradygaster#335), McManus (bradygaster#338), Nate (bradygaster#339).

Changes:
- Session log: 2026-02-23T2228-phase3-launch.md
- Merged decision: P1 UX polish patterns → decisions.md
- Deleted inbox file: cheritto-p1-ux-polish.md

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* style: copy polish — human, fun, action-oriented (closes bradygaster#338)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat: rich progress indicators for agent work (closes bradygaster#335)

- AgentPanel: rich status line with (working/streaming, elapsed) — activity hint
- MessageStream: activity feed showing per-agent actions (📋 Keaton is reading file...)
- AgentSession type: added activityHint field for per-agent activity tracking
- SessionRegistry: updateActivityHint() method for pipeline integration
- App.tsx: agentActivities state + setAgentActivity in ShellApi
- shell/index.ts: wires tool_call events into per-agent activity hints
- 11 new tests in test/repl-ux.test.ts section 9

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs: update cheritto history + decision doc for bradygaster#335

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
style: copy polish — human, fun, action-oriented (bradygaster#358)
19d0e15 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
feat: accessibility hardening and guidelines (bradygaster#359)
18753ae 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 3 launch

Session: 2026-02-23T2228-phase3-launch
Phase 2 complete with P1 UX polish (PR bradygaster#356 merged).
Phase 3 Wave 1 launched: Cheritto (bradygaster#335), McManus (bradygaster#338), Nate (bradygaster#339).

Changes:
- Session log: 2026-02-23T2228-phase3-launch.md
- Merged decision: P1 UX polish patterns → decisions.md
- Deleted inbox file: cheritto-p1-ux-polish.md

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* style: copy polish — human, fun, action-oriented (closes bradygaster#338)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat: accessibility hardening and guidelines (closes bradygaster#339)

- NO_COLOR/TERM=dumb compliance across all shell components
- isNoColor() utility in terminal.ts with noColor in TerminalCapabilities
- AgentPanel: static dot, [Active]/[Error] text labels in monochrome
- ThinkingIndicator: static dots, no color cycling, emoji prefix
- InputPrompt: static [working...], bold cursor
- MessageStream/App: all color props gated on isNoColor()
- docs/accessibility.md: keyboard shortcuts, NO_COLOR matrix, contrast guide
- Decision record: .squad/decisions/inbox/nate-a11y-hardening.md

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
feat: terminal adaptivity 40-120 col range (bradygaster#360)
2b61df2 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat: terminal adaptivity 40-120 col range (closes bradygaster#336)

- Add useTerminalWidth() hook and getTerminalWidth() to terminal.ts
- AgentPanel: compact single-line layout at ≤60 cols, hint truncation at <100
- App.tsx: welcome banner scales from minimal (≤60) to full (≥100)
- InputPrompt: prompt shrinks to 'sq>' at narrow widths (<60)
- MessageStream: separator uses reactive terminal width
- commands.ts: /help shows compact single-column at <80 cols

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
feat: breathtaking first-run wow moment (bradygaster#362)
245ad0e 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 3 Wave 2 merged, final wave launch

Changes:
- Merged 5 decisions from inbox to decisions.md (progress indicators, adaptivity, copy polish, a11y, animations)
- Created session log 2026-02-23T2300-phase3-wave2-merged.md
- Wave 2 PRs bradygaster#360–361 merged with merged conflict resolution
- Wave 3 final launch: Cheritto working on bradygaster#340 P2 and bradygaster#341 wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat: breathtaking first-run wow moment (closes bradygaster#341)

Init ceremony:
- Typewriter reveal of 'Let's build your team' at start
- Staggered landmark reveal (workspace, skills, workflows, identity, agent prompt)
- Celebration: 'Your team is ready. Run squad to start.'
- Respects NO_COLOR (static text, no animation)

First-launch experience:
- .first-run marker written by init, consumed on first shell launch
- Guided prompt: 'Your squad is assembled. Try: @keaton what should we build first?'
- isFirstRun flag in WelcomeData drives conditional rendering

7 new tests in repl-ux.test.ts sections 11 (init ceremony + first-launch)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs: update Cheritto history and decision for wow moment (bradygaster#341)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
test: E2E integration tests for REPL flow (bradygaster#379)
804ab7f 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* test: E2E integration tests for REPL and multi-agent coordination (closes bradygaster#372, closes bradygaster#373)

Adds 15 integration tests covering the interactive REPL pipeline and
multi-agent session coordination — areas that previously had zero
end-to-end coverage.

Tests:
- Full REPL round-trip: user input -> parseInput -> dispatch -> response render
- @agent direct message routing with correct agent targeting
- /help and /status slash commands without SDK dispatch
- Error recovery: dispatch failure + no-SDK-connected states
- Multi-agent session tracking: registration, concurrent status, error cleanup
- Fan-out dispatch: concurrent multi-agent with error isolation
- Input parsing integration with registered agent lists

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* test: hostile input, SDK failure, and stress tests (closes bradygaster#376, closes bradygaster#377, closes bradygaster#378)

- Wire 67-string nasty-inputs corpus into parseInput, executeCommand, and MessageStream rendering tests
- Add SDK failure scenario tests: ghost response, throws, timeouts, error events, malformed data
- Add stress/boundary tests: 1000 messages, rapid dispatch, 1MB inputs, concurrent sessions, MemoryManager limits
- 62 new tests across 3 files, all passing

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs: update breedan history with E2E integration test session

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
test: human journey E2E tests - 7 real user scenarios, 39 assertions (b…
ae4a212 
…radygaster#408)

* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* test: add human journey E2E tests for 7 real user scenarios

Test the actual human experience, not internal modules. Each describe
block maps to a filed GitHub issue and simulates what a real person
does when they encounter Squad for the first time and beyond.

Journeys tested (39 assertions):
1. I just installed this — squad init creates .squad/, shows ceremony,
   gives clear next step, writes first-run marker, exits cleanly
2. My first conversation — welcome banner shows agent roster with
   names/roles/emojis, project description, focus area, first-run
   ceremony is one-time only
3. I'm waiting and getting anxious — ThinkingIndicator appears
   immediately, shows elapsed time, activity hints override default,
   NO_COLOR mode works
4. Something went wrong — friendly error messages with remediation
   tips, no stack traces leaked, whitespace input handled gracefully
5. I want to talk to a specific agent — @agent routing, case-
   insensitive matching, comma syntax, unknown names fall to
   coordinator
6. I'm a power user now — /help, /status, /agents, /history, /quit
   all return useful output, unknown commands guide to /help
7. I came back the next day — state persists, first-run consumed,
   team still loaded, focus area survives between sessions

Closes bradygaster#383, bradygaster#384, bradygaster#385, bradygaster#386, bradygaster#394, bradygaster#396, bradygaster#398

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Copy polish: warmer, human-centered messages across CLI

Every message now feels helpful, not corporate. Changes include:

- cli-entry.ts: 'Your AI team, always ready' instead of generic desc
- commands.ts: Status output changed from 'Root/Active now' to 'Home/Active'
- Help text: 'Check your team' instead of 'Show which squad'
- InputPrompt: 'Message or @agent' instead of corporate placeholder
- App.tsx: 'Your team is assembled. Let's try:' (warmer tone)
- init.ts: 'Your team is assembled' instead of 'ready'
- ThinkingIndicator: Dynamic thinking phrases (Processing, Working)
- All command help: Casual verbs (Who's here, Fresh screen, Leave)

Tests updated to match new output strings. All 125+ CLI tests pass.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
Product Love: First 10 seconds that make you want the next 10 minutes (…
545b6e3 
…bradygaster#414)

* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: product love — first 10 seconds that make you want the next 10 minutes

Walked through the full first-time user experience and fixed every moment
that wasn't delightful:

**\/clear\ actually clears (bradygaster#400)**
- Was sending raw ANSI escape as message content (no-op in Ink)
- Now resets message array to empty — genuine fresh slate
- Added \clear?: boolean\ to CommandResult interface

**Natural language routing revealed (bradygaster#402)**
- The coordinator auto-routing is the WOW feature — was completely hidden
- Welcome banner hints: 'Just type · @agent to direct · /help · Ctrl+C exit'
- /help explains: 'Just type naturally — the coordinator routes it'
- First-run prompt: adds 'Or just type naturally' hint
- Input placeholder: 'Type anything or @agent...'

**First-run hint layout (bradygaster#404)**
- Was horizontal Box that broke on narrow terminals
- Now vertical flexDirection=column with breathing room

**Exit message consistency (bradygaster#405)**
- Replaced wave emoji with diamond mark (matches P2 emoji removal)

**Welcome roster wrapping (bradygaster#407)**
- Agents were one dense string that wrapped mid-name
- Now individual Ink elements with flex-wrap for clean word-boundary wrapping

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
fix: correct ThinkingIndicator test assertion (bradygaster#415)
8086fec 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: correct ThinkingIndicator assertion to match component behavior

The ThinkingIndicator renders empty string when isThinking=false,
not 'No agents active'. Fix the test assertion.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
Fix bradygaster#422: Add context to thinking spinner (bradygaster#435)
8e1c9f3 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix bradygaster#427: Add immediate shell launch indicator

Adds 'Loading Squad shell...' message at start of runShell() to eliminate
2-4 second launch dead air. Message clears once Ink mounts.

Users now see feedback within 100ms instead of staring at blank terminal.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
Fix bradygaster#422: Add context to thinking spinner (bradygaster#436)
41c214f 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix bradygaster#422: Add context to thinking spinner

Changed ThinkingIndicator default label from 'Thinking...' to
'Routing to agent...' to give users meaningful feedback during
SDK connection and initial routing phases.

When activityHint is provided (e.g., 'Keaton thinking...'), it
still takes priority. The new default eliminates the 'is it broken?'
anxiety during the 3-5 second cold connection wait.

Updated tests to reflect new default label.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs: Update Marquez history with bradygaster#422 resolution

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix bradygaster#420/bradygaster#425: Add immediate SDK connection feedback

Before this fix, the first message sent to the REPL had 2-7 seconds of
dead air while createSession() blocked on SDK connection. Users thought
the shell was hung.

Changes:
- Set 'Connecting to SDK...' hint BEFORE createSession() in dispatchToCoordinator
- Set 'Connecting to <agent>...' hint BEFORE createSession() in dispatchToAgent
- Use setImmediate to give React a tick to render before blocking
- Update hint to 'Routing...' or 'thinking...' after connection completes

The ThinkingIndicator now displays immediately, eliminating perceived hang.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
Fix bradygaster#420/bradygaster#425: Add SDK connection feedback (bra…
6f8258a 
…dygaster#437)

* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix bradygaster#420/bradygaster#425: Add immediate SDK connection feedback

Before this fix, the first message sent to the REPL had 2-7 seconds of
dead air while createSession() blocked on SDK connection. Users thought
the shell was hung.

Changes:
- Set 'Connecting to SDK...' hint BEFORE createSession() in dispatchToCoordinator
- Set 'Connecting to <agent>...' hint BEFORE createSession() in dispatchToAgent
- Use setImmediate to give React a tick to render before blocking
- Update hint to 'Routing...' or 'thinking...' after connection completes

The ThinkingIndicator now displays immediately, eliminating perceived hang.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
Fix bradygaster#432: Show retry warnings earlier (bradygaster#441)
51032ac 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix bradygaster#432: Show ghost response retry warnings earlier

Ghost response retry now shows clear progress:
- First retry: 'Empty response detected. Retrying... (attempt 2/4)'
- Subsequent retries show current/total: (attempt 3/4), (attempt 4/4)
- Final error shows accurate total: 'after 4 attempts'

Previously showed confusing counts that didn't match reality.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
fix: update test assertions to match new UX copy (bradygaster#443)
a6baa37 
* chore(squad): Phase 2 launch — thinking feedback, P0 bugs, dual telemetry

Phase 1 complete: 5 issues closed (bradygaster#325, bradygaster#326, bradygaster#327, bradygaster#328, bradygaster#329), 5 PRs merged.
Phase 2 launched with Cheritto (thinking feedback), Hockney (P0 bugs), Saul (dual telemetry).
Decision inbox merged and archived.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Phase 2 Wave 1 merged, Wave 2 launched

Session: 2026-02-23T2145-phase2-wave2
Phase 2 Wave 1 complete (PRs bradygaster#351, bradygaster#352, bradygaster#353 merged).
Wave 2 launched: Cheritto on ghost response detection (bradygaster#332), Hockney on error hardening (bradygaster#334).

Changes:
- Session log created: 2026-02-23T2145-phase2-wave2.md
- Merged 3 inbox decisions (Cheritto, Hockney, Saul)
- Deleted inbox files post-merge

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): Epic bradygaster#323 complete — all phases shipped 🎉

All 3 phases delivered:
- Phase 1 (Testing Wave): 6 issues closed
- Phase 2 (Improvement): 6 issues closed
- Phase 3 (Breathtaking): 7 issues closed
- 17 PRs merged, 19 issues closed total

Session log: 2026-02-23T2320-epic-complete.md
Decisions merged from inbox: P2 UX Polish, first-run wow moment

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* hostile QA: end-to-end quality assessment — 10 findings, 4 HIGH severity

Candid assessment requested by Brady. Traced every code path in cli-entry.ts,
shell/index.ts, shell/commands.ts, App.tsx, coordinator.ts, spawn.ts, and the
SDK adapter client.

Key findings:
- Dead sessions never evicted from agentSessions Map after connection drop
- No React ErrorBoundary — any render throw kills the shell
- Nasty-inputs corpus (95 strings) is never imported by any test
- No SIGTERM handler in interactive shell
- MemoryManager exported but never instantiated (dead code)
- Single streaming content slot clobbers multi-agent output
- User input silently dropped during processing (no type-ahead buffer)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): quality review findings — 7 issues filed

Quality audit complete: 5 agents assessed CLI across testing, coverage, stability, accessibility, UX.
Results: 4 P0 blockers (bradygaster#365–bradygaster#368), 3 P1 items (bradygaster#369–bradygaster#371).
Blocking: Waingro dead sessions, ErrorBoundary, dropped input; Marquez help text consistency.

Changes:
- Logged session summary to .squad/log/2026-02-24T0205-quality-review-complete.md
- Updated .squad/identity/now.md with quality review findings and new issue numbers

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): merge decision — Marquez UX audit findings

Quality assessment merged from inbox (Grade B): 11 improvements (3 P0, 4 P1, 4 P2). help text, stub commands, vocabulary, separators, roster.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore(squad): test sprint launch

Session: 2026-02-24T0210-test-sprint
Changes:
- Logged test sprint: 5 agents, 7+ issues
- Branches: P0 fixes, stale tests, E2E, hostile/SDK, A11y

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: update test assertions to match new UX copy

ThinkingIndicator now says 'Routing to agent...' instead of 'Thinking'.
Help uses progressive disclosure — /clear moved to /help full.
--help first 5 lines now show 'route/agent' instead of 'usage'.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ab3y pushed a commit to Ab3y/squad that referenced this pull request Mar 10, 2026
@bradygaster
docs(ai-team): Summarize history files to Core Context (under 12KB th…
8d3043c 
…reshold)

- Cheritto: Archive Feb 23–26 work (bradygaster#325–bradygaster#446) into Core Context, keep recent work visible
- Kovash: Archive Feb 23–24 diagnostics & fixes (bradygaster#437–bradygaster#442) into Core Context
- Fortier: Archive Feb 21–22 SDK/OTel/Wave 2 work into Core Context

All three files now under 12KB threshold. Readability improved for future sessions.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@EmmittJ EmmittJ changed the base branch from dev to main March 10, 2026 13:25
@EmmittJ EmmittJ force-pushed the EmmittJ/297-skill-script-loader branch from cf5cb84 to ab5183a Compare March 10, 2026 13:28
EmmittJ and others added 2 commits March 10, 2026 06:34
@EmmittJ
feat(skills): add SkillScriptLoader, handler types, and applySkillHan…
9c0cb4a 
…dlers

Implements the executable skill handler layer from PR bradygaster#215 design proposal.
Inspired by bradygaster#297.

- handler-types.ts: full type system for skill-script model (SkillHandler<T>,
  HandlerLifecycle, concern interfaces, defineHandler() helper)
- skill-script-loader.ts: SkillScriptLoader class, resolveSkillPath(),
  wrapSkillHandler(), CONCERN_TOOL_MAP
- applySkillHandlers(): ToolRegistry integration for runtime handler replacement
- test/skill-script-loader.test.ts: 33 tests covering load, execute, error paths

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@EmmittJ
refactor(skill-script-loader): replace CONCERN_TOOL_MAP with director…
669b7a7 
…y scan

Drop the hardcoded concern-to-tool-name map. SkillScriptLoader.load()
now scans scripts/ for .js files and derives tool names from filenames:
create_issue.js -> squad_create_issue. The concern parameter is removed
from load() — the scripts/ directory is the source of truth.

- Remove CONCERN_TOOL_MAP and Concern/ConcernMap imports
- load() now uses readdirSync to discover handler files
- Tool name derivation: 'squad_' + stem (e.g. create_issue.js -> squad_create_issue)
- Update tests: remove concern arg from all .load() calls, simplify
  multi-concern test to 'discovers scripts for any naming convention'
- Update skill-system-design.md: bidirectional convention docs,
  directory-driven scanning in loader section and startup sequence,
  doctor validation updated to remove CONCERN_TOOL_MAP reference

All 33 skill-script-loader tests pass.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@EmmittJ EmmittJ force-pushed the EmmittJ/297-skill-script-loader branch from ab5183a to 669b7a7 Compare March 10, 2026 13:34
@EmmittJ
feat(skills): OTel instrumentation for skill handlers, resolveSkillPa…
c9a41c5 
…th DRY refactor, backendConfig security docs

- Add wrapSkillHandlerWithSpan() in skill-script-loader.ts — tools exit the
  loader already instrumented; closes the 'Known Limitation: OTel bypass' on
  applySkillHandlers. Uses squad.skill.* span namespace (squad.skill.call /
  squad.skill.result / squad.skill.error) to distinguish from squad.tool.*
  spans emitted by defineTool().
- Revert tools/index.ts OTel to original inline form — skills and tools have
  different signatures and concerns; implementations are intentionally separate.
- Update applySkillHandlers JSDoc — no more Known Limitation comment.
- Refactor resolveSkillPath — extract realOrLogical() and isContained() inner
  helpers to remove duplicated realpathSync+fallback and containment logic
  across the three path branches.
- Update backendConfig @warning in skill-script-loader.ts and handler-types.ts
  — clarifies this config is committed to the repo; do not store credentials,
  tokens, or secrets here.
@bradygaster bradygaster mentioned this pull request Mar 12, 2026
Closed
14 tasks
@bradygaster
Copy link
Copy Markdown
Owner

bradygaster commented Mar 15, 2026

Hey @EmmittJ — thank you so much for this work! Your SkillScriptLoader implementation has already been merged to dev (via commit ba8d8f7), so this PR is now covered. The skill handler system you designed is live and part of the runtime. Really appreciate the contribution — it's made a meaningful difference to how skills work in Squad. Closing this PR since the code is already in. 🎉

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@EmmittJ @bradygaster