fvtt-world-cli 1.0.0 → 1.1.1

package/README.md

@@ -2,24 +2,24 @@
 
 [![Foundry VTT v13–v14](https://img.shields.io/badge/Foundry%20VTT-v13%E2%80%93v14-ff6400)](https://foundryvtt.com)
 
-Tell your AI agent what should change in your Foundry VTT world, and it happens — in the live
+Tell your AI agent what should change in your Foundry VTT world, and it happens in the live
 world, validated by Foundry, visible to your players immediately.
 
 ```text
-You → agent:  “The scimitar in Valeros's inventory should burn targets on hit —
-               add a Flaming effect to it.”
+You → agent:  "The scimitar in Valeros's inventory should burn targets on hit.
+               Add a Flaming effect to it."
 
 agent → fvtt-world-cli:
               actor list --name Valeros            find the character
               actor item list --name scimitar      find the sword in their inventory
               actor item effect create …           add the effect to it
 
-agent → you:  “Done — Flaming effect added, already live in the world.”
+agent → you:  "Done. Flaming effect added, already live in the world."
 ```
 
 Modern agents already understand requests like that. What they have lacked is a safe way into
 Foundry: driving the browser UI is brittle and token consuming, and editing world files on disk
-bypasses everything Foundry does to keep a world consistent. fvtt-world-cli is the missing bridge —
+bypasses everything Foundry does to keep a world consistent. fvtt-world-cli is the missing bridge,
 a command line wired into your open GM session that performs every change through Foundry's own
 APIs, exactly as if a GM had made it in the UI.
 
@@ -34,18 +34,18 @@ scenes down to individual tokens, walls, and lights, and pull content in from co
 
 A few asks it handles end to end:
 
-- “Sort the journals into folders by location.”
-- “Dim every light in the tavern scene down to torchlight.”
-- “Find every NPC that still has no portrait and list them.”
-- “Turn the bestiary goblin into a flying one that throws dynamite for 2d6 damage, and add it to
-  the scene.”
+- "Sort the journals into folders by location."
+- "Dim every light in the tavern scene down to torchlight."
+- "Find every NPC that still has no portrait and list them."
+- "Turn the bestiary goblin into a flying one that throws dynamite for 2d6 damage, and add it to
+  the scene."
 
 [Commands](docs/commands.md) maps the full surface.
 
 ## Setting up
 
 The simplest setup is one step: point your AI agent at this repository and ask it to set
-fvtt-world-cli up. Then follow its lead — it will most likely ask you to install the module in
+fvtt-world-cli up. Then follow its lead. It will most likely ask you to install the module in
 Foundry and click *Pair* there, and it handles the rest itself.
 
 ### Manual setup
@@ -75,14 +75,14 @@ agent works with the world.
 
 4. Pair the browser: choose *Pair* in the module's Authorization window, reachable from the
    *World CLI* group in the scene controls or from the module's settings. Then approve the request
-   from the terminal — it shows the requesting origin, world, GM, and browser, and asks for a yes
+   from the terminal. It shows the requesting origin, world, GM, and browser, and asks for a yes
    or no:
 
    ```bash
    fvtt-world-cli auth
    ```
 
-5. Optionally, install the packaged skill into your AI agent — the next section explains what it
+5. Optionally, install the packaged skill into your AI agent; the next section explains what it
    does:
 
    ```bash
@@ -100,8 +100,9 @@ The package ships `foundry-world-editor`, an installable [Agent Skill](docs/skil
 an agent to drive the CLI safely. It follows the open Agent Skills standard, so the same skill
 works in Claude Code, Codex, and other agents that read `SKILL.md` files.
 
-An installed skill stays current on its own: updating the package refreshes it. A copy you have
-edited locally is never replaced automatically, so your own instructions survive updates.
+An installed skill stays current on its own: after a package update, the bridge daemon refreshes
+it at startup. A copy you have edited locally is never replaced automatically, so your own
+instructions survive updates.
 
 ## Built to be trusted with a live world
 
@@ -116,6 +117,8 @@ authority; [Security](docs/security.md) covers the boundaries in full. In short:
   the UI uses, under the GM's permissions.
 - Any change can be previewed before it happens: a global `--dry-run` flag runs the same
   validation and guards as a real call and stops before mutation.
+- Every command has a permission in the active Foundry client: run, ask the GM, or refuse.
+  Destructive commands ask the GM by default.
 - There is no arbitrary-code path: commands are typed and validated on both sides of the transport,
   and executable content such as scripted region behaviors is blocked on every write route.
 - File access is confined to the active world's managed assets and always excludes its manifest,

package/bin/fvtt-world-cli.js

@@ -31,7 +31,19 @@ if (existsSync(compiledEntry) && process.env.FVTT_WORLD_CLI_FORCE_SRC !== "1") {
     stdio: "inherit"
   });
 
+  /** @type {NodeJS.Signals[]} */
+  const relayedSignals = ["SIGINT", "SIGTERM", "SIGHUP"];
+  for (const signal of relayedSignals) {
+    process.on(signal, () => {
+      child.kill(signal);
+    });
+  }
+
   child.on("exit", (code, signal) => {
+    for (const relayed of relayedSignals) {
+      process.removeAllListeners(relayed);
+    }
+
     if (signal) {
       process.kill(process.pid, signal);
       return;