@rallycry/conveyor-agent 11.0.13 → 11.0.14

package/dist/{chunk-SKMNL6C2.js → chunk-LZWCJHVN.js}

@@ -6956,7 +6956,16 @@ var CodexTuiAdapter = class {
       cli_auth_credentials_store: "file",
       sqlite_home: codexSqliteHome(this.env),
       approval_policy: "never",
-      sandbox_mode: input.options.permissionMode === "plan" ? "read-only" : "danger-full-access",
+      // Always full access, plan mode included. Codex's own read-only mode is
+      // bubblewrap on Linux, which needs user namespaces the GKE pod spec
+      // forbids (seccomp RuntimeDefault, no unprivileged userns): every shell
+      // command died with "bwrap: No permissions to create a new namespace"
+      // before it started. A non-full-access sandbox under approval_policy
+      // "never" also rejects every mutating MCP call ("MCP tool call requires
+      // approval"), so the agent could not even report the blocker. Plan
+      // sessions stay read-only the way Claude and OpenCode plan sessions do:
+      // reader GitHub token, mode prompt, disposable pod.
+      sandbox_mode: "danger-full-access",
       developer_instructions: [
         input.options.appendSystemPrompt,
         "Conveyor workflow skills are stored in .claude/skills in the workspace. When instructed to run a /skill-name skill, read .claude/skills/skill-name/SKILL.md and follow its instructions. This is a file-based workflow, not a native slash command."

package/dist/cli.js

@@ -27,7 +27,7 @@ import {
   resolveTuiKindFromEnv,
   runUsageProbe,
   sampleKeyUsage
-} from "./chunk-SKMNL6C2.js";
+} from "./chunk-LZWCJHVN.js";
 import "./chunk-SR66HQKB.js";
 import {
   inheritedEnv,

package/dist/index.js

@@ -1,7 +1,7 @@
 import {
   SessionRunner,
   unshallowRepo
-} from "./chunk-SKMNL6C2.js";
+} from "./chunk-LZWCJHVN.js";
 import "./chunk-SR66HQKB.js";
 import "./chunk-JG3VMWRQ.js";
 import {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rallycry/conveyor-agent",
-  "version": "11.0.13",
+  "version": "11.0.14",
   "description": "Conveyor Agent Runner v10 - PTY harness for the task chat (SDK harness for audit/project-chat). Agent-as-User architecture with BaseService patterns. Works locally too.",
   "keywords": [
     "agent",