@cjhyy/code-shell-capability-coding 0.8.0 → 0.8.2

package/dist/external-runtimes/claude-code/mcp-config.d.ts

@@ -49,17 +49,15 @@ export declare function writeClaudeMcpConfigFile(options: ClaudeMcpConfigOptions
  *
  * Safety therefore rests on ONE property, enforced below rather than assumed:
  * every emitted name is `mcp__<server>__<tool>`, which routes to CodeShell's own
- * `ToolExecutor` and exposure policy. A bare `Bash` or `Write` reaching this list
- * would grant the untrusted runtime an unapproved built-in tool, and the exposure
- * allowlist would not help — it governs the MCP surface, not Claude Code's own
- * tools.
+ * `ToolExecutor` and exposure policy. CodeShell deliberately exposes governed
+ * alternatives named `Read`, `Bash`, `Write`, and `Edit`; after qualification
+ * these become `mcp__codeshell_tools__Read`, etc. They do not pre-approve Claude
+ * Code's bare built-ins with the same names.
  *
- * Note what is NOT done here: auto-prefixing. Turning `"Bash"` into
- * `mcp__codeshell_tools__Bash` would make every input "valid" and defeat the
- * check — the first version of this function did exactly that. A CodeShell tool
- * name is a plain identifier, so it must be a plain identifier, and anything that
- * looks like a built-in-tool pattern (`Bash(git *)`, `*`) or is already qualified
- * for a different server is a programming error.
+ * The function accepts only plain tool identifiers sourced from the exposed
+ * CodeShell registry, then adds the MCP namespace itself. Anything that looks
+ * like a Claude permission pattern (`Bash(git *)`, `*`) or is already qualified
+ * for another server is a programming error and is rejected.
  */
 export declare function claudeAllowedToolNames(toolNames: Iterable<string>, serverName?: string): string[];
 /**

package/dist/external-runtimes/claude-code/mcp-config.js

@@ -24,30 +24,6 @@ import { join } from "node:path";
 export const CLAUDE_MCP_SERVER_NAME = "codeshell_tools";
 /** `mcp__<server>__<tool>` — the only shape that may reach `--allowed-tools`. */
 const MCP_TOOL_NAME = /^mcp__[a-z0-9_]+__[A-Za-z0-9_]+$/;
-/**
- * Claude Code's own built-in tools. They share the plain-identifier shape with
- * CodeShell tool names, so shape validation alone cannot separate them — and
- * pre-approving one of these grants the untrusted runtime an unapproved built-in.
- * Over-inclusive on purpose: a false positive is a loud error at a call site, a
- * false negative is a silent capability grant.
- */
-const CLAUDE_BUILTIN_TOOL_NAMES = new Set([
-    "Agent",
-    "Bash",
-    "BashOutput",
-    "Edit",
-    "ExitPlanMode",
-    "Glob",
-    "Grep",
-    "KillShell",
-    "NotebookEdit",
-    "Read",
-    "Task",
-    "TodoWrite",
-    "WebFetch",
-    "WebSearch",
-    "Write",
-]);
 /**
  * The `--mcp-config` payload as an inline JSON string.
  *
@@ -107,17 +83,15 @@ export function writeClaudeMcpConfigFile(options) {
  *
  * Safety therefore rests on ONE property, enforced below rather than assumed:
  * every emitted name is `mcp__<server>__<tool>`, which routes to CodeShell's own
- * `ToolExecutor` and exposure policy. A bare `Bash` or `Write` reaching this list
- * would grant the untrusted runtime an unapproved built-in tool, and the exposure
- * allowlist would not help — it governs the MCP surface, not Claude Code's own
- * tools.
+ * `ToolExecutor` and exposure policy. CodeShell deliberately exposes governed
+ * alternatives named `Read`, `Bash`, `Write`, and `Edit`; after qualification
+ * these become `mcp__codeshell_tools__Read`, etc. They do not pre-approve Claude
+ * Code's bare built-ins with the same names.
  *
- * Note what is NOT done here: auto-prefixing. Turning `"Bash"` into
- * `mcp__codeshell_tools__Bash` would make every input "valid" and defeat the
- * check — the first version of this function did exactly that. A CodeShell tool
- * name is a plain identifier, so it must be a plain identifier, and anything that
- * looks like a built-in-tool pattern (`Bash(git *)`, `*`) or is already qualified
- * for a different server is a programming error.
+ * The function accepts only plain tool identifiers sourced from the exposed
+ * CodeShell registry, then adds the MCP namespace itself. Anything that looks
+ * like a Claude permission pattern (`Bash(git *)`, `*`) or is already qualified
+ * for another server is a programming error and is rejected.
  */
 export function claudeAllowedToolNames(toolNames, serverName = CLAUDE_MCP_SERVER_NAME) {
     const reject = (name, why) => {
@@ -136,11 +110,6 @@ export function claudeAllowedToolNames(toolNames, serverName = CLAUDE_MCP_SERVER
         if (name.startsWith("mcp__")) {
             return reject(name, "already qualified — pass a plain CodeShell tool name");
         }
-        // Claude Code's own built-ins share the identifier shape, so shape alone is
-        // not enough — these must never be pre-approved through this path.
-        if (CLAUDE_BUILTIN_TOOL_NAMES.has(name)) {
-            return reject(name, "this is a Claude Code built-in tool, not a CodeShell tool");
-        }
         const qualified = `mcp__${serverName}__${name}`;
         if (!MCP_TOOL_NAME.test(qualified))
             return reject(name, "does not qualify as an MCP tool name");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cjhyy/code-shell-capability-coding",
-  "version": "0.8.0",
+  "version": "0.8.2",
   "description": "Coding capability pack for the generic code-shell agent core.",
   "type": "module",
   "main": "./dist/index.js",
@@ -39,7 +39,7 @@
     "clean": "node -e \"require('node:fs').rmSync('dist',{recursive:true,force:true})\""
   },
   "dependencies": {
-    "@cjhyy/code-shell-core": "0.8.0"
+    "@cjhyy/code-shell-core": "0.8.2"
   },
   "engines": {
     "node": ">=20.10"