imagegen-smarto 0.1.1 → 0.1.2

package/README.md

@@ -5,8 +5,15 @@ through a custom SmartO OpenAI-compatible VPS relay.
 
 The skill does not call Codex's built-in `image_gen` tool. It places a private
 routing marker in the Responses instructions; the relay adds its
-`image_generation` tool only for requests that load this skill. Ordinary
-non-image requests remain unchanged.
+`image_generation` tool for requests that load this skill, merging it with
+existing function, shell, file, or other tools. Ordinary non-image requests
+remain unchanged.
+
+When the SmartO profile is active, this skill can also replace only the image
+provider inside workflows that normally use `$imagegen`, including the
+official `hatch-pet` workflow. The workflow's prompts, references, QA, and
+file handling remain unchanged; only image generation is routed through
+SmartO.
 
 ## Install with npm
 

package/imagegen-smarto/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: imagegen-smarto
-description: Generate or edit raster images through the user's SmartO OpenAI-compatible VPS relay. Automatically use this skill for natural-language requests to create, draw, generate, or edit an image when the SmartO relay profile is active; do not use it for image analysis or coding tasks, official OpenAI/ChatGPT providers, or the built-in image_gen tool.
+description: Generate or edit raster images through the user's SmartO OpenAI-compatible VPS relay. Automatically use this skill for natural-language requests to create, draw, generate, or edit an image when the SmartO relay profile is active. It is also the drop-in image-generation provider for workflows that normally invoke `$imagegen`, including `hatch-pet`; preserve those workflows and replace only the image-generation execution layer. Do not use it for image analysis or coding tasks, official OpenAI/ChatGPT providers, or the built-in image_gen tool.
 ---
 
 # SmartO relay image generation
@@ -10,6 +10,20 @@ SmartO relay profile is active. It is implicitly invokable, so the user should
 be able to ask for an image in natural language without first writing
 `$imagegen-smarto`.
 
+When another active skill invokes `$imagegen` for a visual job (including the
+official `hatch-pet` skill), treat this skill as the selected image-generation
+provider when the SmartO relay profile is active. Keep the calling skill's
+workflow, prompts, reference-image requirements, QA, and file handling
+unchanged; replace only the provider-specific built-in image-generation call.
+If that skill also loads the official `.system/imagegen/SKILL.md`, retain its
+prompting and output-quality guidance but override only its provider-selection
+rules: do not call the built-in `image_gen` tool and do not use its CLI/API
+fallback. Use the normal Responses request through SmartO instead.
+Every separate image-generation worker request must load this routing
+instruction so the marker below is present in that request's Responses
+`instructions` field. Do not ask the user to type the marker or add it to the
+image prompt.
+
 This skill is a routing signal, not a local image-generation tool:
 
 - Never call the built-in `image_gen` tool or any other native image tool.
@@ -27,14 +41,18 @@ built-in image tool:
 `__CODEX_VPS_IMAGEGEN__`
 
 The relay-side rule should apply only to `/v1/responses` requests whose
-`instructions` contain that marker. It should add this tool when the request
-does not already declare tools:
+`instructions` contain that marker. It must merge the image tool into the
+request's existing `tools` array, preserving every client-declared tool. If
+the request has no `tools`, create the array. Do not skip injection merely
+because the request already contains shell, file, function, or other tools:
 
 ```json
 {"type":"image_generation","action":"auto"}
 ```
 
-It may set `tool_choice` to `auto` only when the client did not provide one.
+Do not add a duplicate if an equivalent `image_generation` tool is already
+present. It may set `tool_choice` to `auto` only when the client did not
+provide one.
 Do not force image generation for every turn. The upstream model decides from
 the natural-language request whether to generate a new image or edit an image
 in context.

package/imagegen-smarto/agents/openai.yaml

@@ -1,7 +1,7 @@
 interface:
   display_name: "SmartO VPS Image Generation"
-  short_description: "Generate images through the SmartO VPS relay"
-  default_prompt: "Use $imagegen-smarto to generate or edit an image through my SmartO VPS relay."
+  short_description: "Generate images through SmartO, including $imagegen workflows"
+  default_prompt: "Use SmartO as the image-generation provider for this workflow, including workflows that normally invoke $imagegen."
 
 policy:
   allow_implicit_invocation: true

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "imagegen-smarto",
-  "version": "0.1.1",
+  "version": "0.1.2",
   "description": "Install the imagegen-smarto Codex skill on Windows and Linux.",
   "bin": {
     "imagegen-smarto": "bin/cli.js"