[FEAT]: Allow Agent Skills to Access Images Attached or Pasted in Chat #3210

New issue

Open

opened 2026-02-28 06:33:34 -05:00 by deekerman · 0 comments

deekerman commented

2026-02-28 06:33:34 -05:00

Owner

Originally created by @LudoE11 on GitHub (Feb 25, 2026).

What would you like to see?

Context

I'm building a custom agent skill for AnythingLLM that edits images using an external API (Replicate) to call an AI Image editing model. The ideal user experience is for users to simply attach or paste an image into the chat and have the agent skill process it automatically based on their prompt (send the input image and the prompt to the API).

Current Situation

Currently, when a user sends a message to an agent skill (e.g., using @agent), only the text content of the message is passed to the skill handler. Any images attached or pasted into the chat are not accessible to the skill—there is no URL, file path, or base64 data provided. As far as I know.

Problem

This limitation means users must manually convert images to base64 or upload them to a specific folder, which adds friction and makes the workflow less user-friendly. The skill cannot process images directly from the chat, which is a common and expected use case for image editing or analysis skills.

Questions

Is there currently any way for agent skills to access images that users attach or paste into the chat?
If not, are there any recommended workarounds for this use case that don't add a lot of user friction?
Are there plans to support passing image data (URL, file path, or base64) to agent skills in the future?

Desired Feature

It would be very useful and helfpuf if you could provide a way for agent skills to access images that users attach or paste into the chat, ideally as a URL, file path, or base64 string, so that skills can process them directly and provide a seamless user experience for image-based workflows.

Thanks in advance for your feedback.

Originally created by @LudoE11 on GitHub (Feb 25, 2026). ### What would you like to see? ## Context I'm building a custom agent skill for AnythingLLM that edits images using an external API (Replicate) to call an AI Image editing model. The ideal user experience is for users to simply attach or paste an image into the chat and have the agent skill process it automatically based on their prompt (send the input image and the prompt to the API). ## Current Situation Currently, when a user sends a message to an agent skill (e.g., using `@agent`), only the text content of the message is passed to the skill handler. Any images attached or pasted into the chat are not accessible to the skill—there is no URL, file path, or base64 data provided. As far as I know. ## Problem This limitation means users must manually convert images to base64 or upload them to a specific folder, which adds friction and makes the workflow less user-friendly. The skill cannot process images directly from the chat, which is a common and expected use case for image editing or analysis skills. ## Questions 1. Is there currently any way for agent skills to access images that users attach or paste into the chat? 2. If not, are there any recommended workarounds for this use case that don't add a lot of user friction? 3. Are there plans to support passing image data (URL, file path, or base64) to agent skills in the future? ## Desired Feature It would be very useful and helfpuf if you could provide a way for agent skills to access images that users attach or paste into the chat, ideally as a URL, file path, or base64 string, so that skills can process them directly and provide a seamless user experience for image-based workflows. Thanks in advance for your feedback.

deekerman added the

enhancement

feature request

labels

2026-02-28 06:33:34 -05:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/anything-llm#3210

No description provided.

Rows
Columns