From #283 (comment) by @brigleb:
Once this lands, the SKILL.md could benefit from a short "Working with Inline Images" section that walks an AI agent through the full loop: fetch a recording's JSON, extract <bc-attachment> URLs from the content HTML (using the new richtext.ExtractAttachments capability), download them to a temp directory with basecamp files download "<storage-url>" --out /tmp/, and then view them with the agent's native file-read tool — which, for multimodal LLMs like Claude and Gemini, renders images directly without needing a browser.
Right now the skill documents the CLI commands but doesn't connect the dots for an agent that can see images but doesn't know it should look for them. A brief recipe would close that gap and prevent agents from silently skipping visual context — mockups, screenshots, annotated designs — that's often the most important part of a Basecamp todo for us.
Two pieces of work here:
-
SKILL.md recipe: Add a "Working with Inline Attachments" section documenting the end-to-end agent workflow (fetch recording JSON → extract <bc-attachment> storage URLs → files download → view with native tool).
-
--download-attachments flag: Consider a flag on show commands (todos, messages, etc.) that automatically extracts and downloads inline attachments in one shot, eliminating the multi-step dance for the most common case.
From #283 (comment) by @brigleb:
Two pieces of work here:
SKILL.md recipe: Add a "Working with Inline Attachments" section documenting the end-to-end agent workflow (fetch recording JSON → extract
<bc-attachment>storage URLs →files download→ view with native tool).--download-attachmentsflag: Consider a flag onshowcommands (todos, messages, etc.) that automatically extracts and downloads inline attachments in one shot, eliminating the multi-step dance for the most common case.