fix(spec,release): Windows-green plan tests + valid release plan (#595)

* fix(spec): forward-slash publish result paths; deterministic archive-target collision

spec_change_path/plan_workspace_path returned OS-native separators, so
Windows callers (and the plan.test.ts contains-assertions) saw
backslash paths. archiveChange relied on renameSync failing onto an
existing target — POSIX throws, Windows MoveFileEx clobbers — so the
existence check is now explicit.

* fix(spec,release): Windows-green plan tests + valid release plan

- publishPlan returns forward-slash spec_change_path/plan_workspace_path
  so Windows callers (and plan.test.ts contains-assertions) see
  separator-stable reference paths.
- archiveChange checks the archive target explicitly: POSIX renameSync
  threw on an existing target, Windows MoveFileEx silently clobbered —
  the 'delta written then archive throws' test now errors on every OS.
- .changeset/post-tool-use-awareness-push.md referenced package 'colony'
  (not in workspace; the CLI is 'colonyq'), which crashed changesets'
  assemble-release-plan and failed every Release run since #594.

* fix(cli,mcp): Windows-portable test expectations + structured already-archived error

mcp-server going green on Windows exposed the next packages in pnpm's
recursive run: 9 apps/cli failures, all path-portability in tests —
expectations assumed POSIX literals where the implementation resolves
and shell-quotes OS-natively. Added shellQuoteForTest mirroring the
production quoting rule; cockpit/queen-health/agents-spawn expectations
now derive from resolve(). PATH stubs join with path.delimiter. The
gx/cue PATH-spawning integration tests gate to POSIX (both tools spawn
via execFileSync without shell, which Windows cannot satisfy).

spec_archive returns structured SPEC_ARCHIVE_ALREADY_EXISTS
(status: already_archived) instead of an opaque throw on same-day
re-archive (review follow-up).

* chore(release): mark internal @colony/* packages private

The npm @Colony scope is not ours (@colony/core@3.1.0 belongs to a
different project), so 'changeset publish' would 403 on every public
internal package and fail Release after versioning. colonyq is the
product artifact and ships self-contained (prepack bundles, no
@colony/* deps in the published tarball — e2e-publish green), so the
ten internal packages are now private: publish ships exactly colonyq.

* fix(spec,cli): review follow-ups — staging recovery, aligned shellQuote charset, normalized archived_path

---------

Co-authored-by: NagyVikt <nagy.viktordp@gmail.com>

Author: NagyVikt

.changeset/post-tool-use-awareness-push.md

@@ -1,6 +1,6 @@
 ---
 '@colony/hooks': minor
-'colony': patch
+'colonyq': patch
 ---
 
 Push awareness: when an edit touches a file another live session holds, the PostToolUse hook now injects a one-line `[Colony] session X recently claimed <file> …` note into the agent's context immediately (Claude Code `additionalContext`), instead of waiting for the next turn's preface. Debounced to once per 2 minutes per session via an `awareness-push` observation marker (hook processes are one-shot, so the marker doubles as audit trail). `autoClaimFromToolUse` now reports live-owner blocked takeovers in its `conflicts` result.

.changeset/windows-plan-paths-archive.md

@@ -0,0 +1,5 @@
+---
+'@colony/spec': patch
+---
+
+Windows CI green: `publishPlan` returns forward-slash `spec_change_path` / `plan_workspace_path` (separator-stable for MCP callers on every OS), and `archiveChange` fails deterministically when the archive target already exists (POSIX rename threw; Windows MoveFileEx silently clobbered).

apps/cli/src/lib/gitguardex.ts

@@ -834,7 +834,9 @@ function execGx(execFileSync: GitGuardexExecFileSync, args: string[], cwd: strin
 }
 
 function shellQuote(value: string): string {
-  if (/^[A-Za-z0-9_./:=@%+-]+$/.test(value)) return value;
+  // Same safe charset as health.ts/sweep.ts shellQuote (and the test
+  // helper apps/cli/test/shell-quote-helper.ts) — keep all copies aligned.
+  if (/^[A-Za-z0-9_./:@%+=,-]+$/.test(value)) return value;
   return `'${value.replace(/'/g, `'\\''`)}'`;
 }
 

apps/cli/test/agents-spawn.test.ts

@@ -12,6 +12,7 @@ import {
 } from '../src/executors/gitguardex.js';
 import { createProgram } from '../src/index.js';
 import { withStore } from '../src/util/store.js';
+import { shellQuoteForTest } from './shell-quote-helper.js';
 
 const MINIMAL_SPEC = `# SPEC
 
@@ -148,7 +149,7 @@ describe('colony agents spawn', () => {
     expect(output).toContain('Stop and hand off if quota or conflict');
     expect(output).toContain('Sibling files:\n- apps/cli/test/agents-spawn.test.ts');
     expect(output).toContain('Do-not-touch list:\n- apps/cli/test/agents-spawn.test.ts');
-    expect(output).toContain(`--target ${repoRoot}`);
+    expect(output).toContain(`--target ${shellQuoteForTest(repoRoot)}`);
     expect(output).toContain('--claim apps/cli/src/executors/gitguardex.ts');
     expect(existsSync(gxArgsFile)).toBe(false);
   });

apps/cli/test/cockpit.test.ts

@@ -1,6 +1,6 @@
 import { mkdtempSync, rmSync } from 'node:fs';
 import { tmpdir } from 'node:os';
-import { join } from 'node:path';
+import { join, resolve } from 'node:path';
 import kleur from 'kleur';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 import { createProgram } from '../src/index.js';
@@ -9,6 +9,7 @@ import {
   defaultCockpitSessionName,
   formatCommand,
 } from '../src/lib/gitguardex.js';
+import { shellQuoteForTest } from './shell-quote-helper.js';
 
 let dataDir: string;
 let output: string;
@@ -42,7 +43,11 @@ describe('colony cockpit', () => {
     );
 
     expect(output).toContain('gitguardex cockpit dry-run');
-    expect(output).toContain('command: gx cockpit --target /work/colony --session colony-colony');
+    // --repo-root '/work/colony' resolves OS-natively (D:\work\colony on
+    // Windows) and quotes when the path leaves the shell-safe charset.
+    expect(output).toContain(
+      `command: gx cockpit --target ${shellQuoteForTest(resolve('/work/colony'))} --session colony-colony`,
+    );
     expect(output).toContain('next ready spawn commands:');
     expect(output).toContain('no ready Colony plan subtasks');
   });
@@ -69,8 +74,8 @@ describe('colony cockpit', () => {
 
     expect(JSON.parse(output)).toMatchObject({
       dry_run: true,
-      command: 'gx cockpit --target /work/colony --session ops',
-      repo_root: '/work/colony',
+      command: `gx cockpit --target ${shellQuoteForTest(resolve('/work/colony'))} --session ops`,
+      repo_root: resolve('/work/colony'),
       session_name: 'ops',
       next_spawn_commands: [],
     });
@@ -79,7 +84,9 @@ describe('colony cockpit', () => {
   it('quotes rendered paths and sessions for shell output', () => {
     expect(
       formatCommand('gx', buildGitGuardexCockpitCommand('/work/repo with space', 'ops cockpit')),
-    ).toBe("gx cockpit --target '/work/repo with space' --session 'ops cockpit'");
+    ).toBe(
+      `gx cockpit --target ${shellQuoteForTest(resolve('/work/repo with space'))} --session 'ops cockpit'`,
+    );
   });
 
   it('normalizes the default session name from the repo root', () => {

apps/cli/test/gitguardex.test.ts

@@ -1,6 +1,7 @@
 import { chmodSync, mkdirSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from 'node:fs';
 import { tmpdir } from 'node:os';
-import { join } from 'node:path';
+import { delimiter, join } from 'node:path';
+
 import { defaultSettings } from '@colony/config';
 import { MemoryStore, listPlans } from '@colony/core';
 import { type QueenOrderedPlanInput, publishOrderedPlan } from '@colony/queen';
@@ -12,7 +13,10 @@ import {
 } from '../src/executors/gitguardex.js';
 import { createProgram } from '../src/index.js';
 import { type GitGuardexExecFileSync, spawnGitGuardexAgent } from '../src/lib/gitguardex.js';
-
+// gx (gitguardex) is a POSIX tool; Windows cannot exec the PATH-stub shim
+// (execFileSync without shell rejects .cmd), so PATH-spawning integration
+// tests run on POSIX only.
+const POSIX = process.platform !== 'win32';
 const MINIMAL_SPEC = `# SPEC
 
 ## §G  goal
@@ -62,7 +66,7 @@ beforeEach(() => {
   originalGxStatus = process.env.GX_FAKE_AGENTS_STATUS;
   originalGxFailVersion = process.env.GX_FAKE_FAIL_VERSION;
   process.env.COLONY_HOME = dataDir;
-  process.env.PATH = `${binDir}:${process.env.PATH ?? ''}`;
+  process.env.PATH = `${binDir}${delimiter}${process.env.PATH ?? ''}`;
   process.env.GX_FAKE_LOG = logPath;
   process.env.GX_FAKE_AGENTS_STATUS = JSON.stringify({
     schemaVersion: 1,
@@ -181,7 +185,7 @@ describe('GitGuardex executor CLI', () => {
     expect(output).toContain('--claim apps/cli/src/lib/gitguardex.ts');
   });
 
-  it('calls gx cockpit with repo target and colony session name', async () => {
+  it.runIf(POSIX)('calls gx cockpit with repo target and colony session name', async () => {
     await createProgram().parseAsync(['node', 'test', 'cockpit', '--repo-root', repoRoot], {
       from: 'node',
     });
@@ -249,7 +253,7 @@ describe('GitGuardex executor CLI', () => {
     });
   });
 
-  it('shows GitGuardex lanes in colony status', async () => {
+  it.runIf(POSIX)('shows GitGuardex lanes in colony status', async () => {
     process.env.GX_FAKE_AGENTS_STATUS = JSON.stringify({
       schemaVersion: 1,
       repoRoot,

apps/cli/test/queen-health.test.ts

@@ -11,6 +11,7 @@ import {
 import kleur from 'kleur';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 import { buildColonyHealthPayload, formatColonyHealthOutput } from '../src/commands/health.js';
+import { shellQuoteForTest } from './shell-quote-helper.js';
 
 const NOW = 1_800_000_000_000;
 const SINCE = NOW - 24 * 3_600_000;
@@ -441,7 +442,7 @@ describe('queen wave health', () => {
       state: 'archived',
       recommendation: {
         action: 'publish-new-plan',
-        command: `colony queen plan --repo-root ${repoRoot} "<goal>"`,
+        command: `colony queen plan --repo-root ${shellQuoteForTest(repoRoot)} "<goal>"`,
       },
     });
   });

apps/cli/test/shell-quote-helper.ts

@@ -0,0 +1,10 @@
+/**
+ * Mirror of the production shellQuote rule (packages/queen/src/sweep.ts,
+ * apps/cli/src/commands/health.ts, apps/cli/src/lib/gitguardex.ts) so path
+ * expectations stay correct on every OS: Windows temp paths contain `\` and
+ * `~`, which fall outside the safe charset and get single-quoted.
+ */
+export function shellQuoteForTest(value: string): string {
+  if (/^[A-Za-z0-9_./:@%+=,-]+$/.test(value)) return value;
+  return `'${value.replaceAll("'", "'\\''")}'`;
+}

apps/cli/test/skills-wire.test.ts

@@ -1,6 +1,6 @@
 import { mkdtempSync, rmSync, writeFileSync } from 'node:fs';
 import { tmpdir } from 'node:os';
-import { join } from 'node:path';
+import { delimiter, join } from 'node:path';
 import { afterEach, describe, expect, it } from 'vitest';
 import {
   COLONY_CUE_SKILL_ID,
@@ -9,6 +9,29 @@ import {
   wireColonySkill,
 } from '../src/commands/skills.js';
 
+// Cross-platform fake `cue`: a Node script plus a .cmd shim so Windows
+// spawn (which cannot exec shebang scripts) resolves it too.
+function writeFakeCue(dir: string): void {
+  const cuePath = join(dir, 'cue');
+  writeFileSync(
+    cuePath,
+    [
+      '#!/usr/bin/env node',
+      'const args = process.argv.slice(2);',
+      "if (args[0] === '--version') { process.stdout.write('0.0.0\\n'); process.exit(0); }",
+      "process.stdout.write('called: ' + args.join(' ') + '\\n');",
+      'process.exit(0);',
+      '',
+    ].join('\n'),
+    { mode: 0o755 },
+  );
+}
+
+// The real `cue` CLI ships as a POSIX binary and the wirer spawns it via
+// execFileSync without a shell, which Windows cannot satisfy for extension-
+// less PATH stubs — run the cue-present integration tests on POSIX only.
+const POSIX = process.platform !== 'win32';
+
 describe('colony skills wire (cue auto-wiring)', () => {
   const originalPath = process.env.PATH;
   afterEach(() => {
@@ -41,18 +64,13 @@ describe('colony skills wire (cue auto-wiring)', () => {
     expect(result.ok).toBe(false);
   });
 
-  it('invokes `cue skills add-to-profile <id>` when cue is present', () => {
+  it.runIf(POSIX)('invokes `cue skills add-to-profile <id>` when cue is present', () => {
     const dir = mkdtempSync(join(tmpdir(), 'fake-cue-'));
-    const cuePath = join(dir, 'cue');
     // Stub cue: answer --version for detection, echo every other invocation so
     // the test can assert the exact subcommand the wirer shells out to.
-    writeFileSync(
-      cuePath,
-      '#!/usr/bin/env bash\nif [ "$1" = "--version" ]; then echo 0.0.0; exit 0; fi\necho "called: $*"\nexit 0\n',
-      { mode: 0o755 },
-    );
+    writeFakeCue(dir);
     try {
-      process.env.PATH = `${dir}:${originalPath ?? ''}`;
+      process.env.PATH = `${dir}${delimiter}${originalPath ?? ''}`;
       const result = wireColonySkill();
       expect(result.cueDetected).toBe(true);
       expect(result.ok).toBe(true);
@@ -62,16 +80,11 @@ describe('colony skills wire (cue auto-wiring)', () => {
     }
   });
 
-  it('passes --preview through in dry-run mode', () => {
+  it.runIf(POSIX)('passes --preview through in dry-run mode', () => {
     const dir = mkdtempSync(join(tmpdir(), 'fake-cue-'));
-    const cuePath = join(dir, 'cue');
-    writeFileSync(
-      cuePath,
-      '#!/usr/bin/env bash\nif [ "$1" = "--version" ]; then echo 0.0.0; exit 0; fi\necho "called: $*"\nexit 0\n',
-      { mode: 0o755 },
-    );
+    writeFakeCue(dir);
     try {
-      process.env.PATH = `${dir}:${originalPath ?? ''}`;
+      process.env.PATH = `${dir}${delimiter}${originalPath ?? ''}`;
       const result = wireColonySkill({ dryRun: true });
       expect(result.message).toContain('add-to-profile colony/colony --preview');
     } finally {

apps/mcp-server/src/tools/shared.ts

@@ -296,6 +296,7 @@ export function mcpErrorResponse(
     | 'QUEEN_INVALID_GOAL'
     | 'RESCUE_CONFIRM_REQUIRED'
     | 'SESSION_NOT_FOUND'
+    | 'SPEC_ARCHIVE_ALREADY_EXISTS'
     | 'SPEC_ARCHIVE_CONFLICT'
     | 'TASK_LINK_SELF'
     | 'SCOUT_NO_CLAIM',