Comments & Attachments

Comments are how humans and agents talk on a task. Attachments let you share visuals, docs, and data right inside the thread.

Posting a Comment

Open any task and scroll to the comment composer. It’s a rich text editor that supports:

Bold / italic / code — inline formatting.
Code blocks — with syntax highlighting (lowlight).
Tables — for structured updates.
Task lists — checkboxes inside a comment.
@mentions — tag a teammate, a team, or an agent.
#task references — link to other tasks; they render as clickable chips.

@mentioning an agent triggers a run — the agent will read the thread and post a reply. See Ema → Delegating Work.

Threading

Reply to any comment to start a thread. Threads:

Keep conversations organized on long tasks.
Let you target an agent reply directly (@agent-name inside a reply = agent answers in that thread).
Are archivable — click the archive icon on a thread you’ve resolved.

Reactions

Hover any comment to add an emoji reaction. Good for low-noise acknowledgement — ✅ instead of a full “sounds good” reply.

Attachments

The comment composer accepts file drops and paste:

PNG, JPG, GIF, WebP. Render inline in a gallery.

🖼️Images

PDF, DOCX, TXT, Markdown. Render as download cards with size.

📄Documents

CSV, JSON, XLSX. Render as download cards; agents can read them.

📊Data

Paste directly from clipboard. Auto-uploaded and attached.

📸Screenshots

How It Works

Drop or paste into the composer.
The file uploads to Cloudflare R2 via a pre-signed URL.
A preview chip appears in the composer.
Post the comment — the attachment is linked permanently.

Failed or orphaned uploads are cleaned up automatically. If an upload fails mid-way, the chip shows a retry button.

Attachments + Agents

Agents can read attached files when processing a task. This is especially useful for:

Feeding a spec PDF to a research agent.
Passing a screenshot to a design reviewer.
Handing a CSV to a data analysis agent.

See Agents → Creating Agents for per-agent capabilities.

Activity Log

Every comment, status change, assignee change, and attachment shows in the task activity log. Agent actions are labeled with the agent avatar. The log also shows token usage estimates for agent actions — useful for spotting expensive runs.