Attach markdown files
Use the Markdown panel to give agents input files to read and declare output files they produce.
Cerevisor scans the workflow's project folder for .md files and lists them in the Markdown panel. You can drag those files onto agents to wire them as inputs or outputs, without copying content into prompts. This is a key part of Cerevisor's file intelligence : the harness uses your declared inputFiles[] and outputFiles[] to detect dependencies between agents, avoid write conflicts, and schedule agents in the right order, without ever injecting file content into another agent's context. The Markdown panel Open the panel from the right sidebar. It shows every .md file under the current workflow's folder (recursive, depth 3). The panel: Lists files grouped by sub-folder. Re-scans automatically when you save the workflow or open a different folder. Has a + New file button at the top: creates a blank .md file in the project folder. Has a resizable left edge: drag to widen (150-520 px). Can be collapsed to a 34 px strip via the header chevron. Drag a file onto an agent Two drop zones on each agent card: INPUT (left) : blue chips. Files the agent reads. OUTPUT (right) : green chips. Files the agent is expected to produce. Drag a file from the panel over an agent. The drop zone you're over highlights. Release to attach. You can also click an empty drop zone to open a file picker. Why input/output files matter Two reasons: 1. The harness schedules agents around files If Agent A declares spec.md as an output and Agent B declares spec.md as an input, Cerevisor knows B depends on A, even if you didn't draw a connection. This influences wave scheduling: B won't start until A finishes. 2. Write conflicts are detected If two agents in the same wave both declare the same file as an output, Cerevisor warns you before running. (Two agents writing the same file at the same time is almost never what you want.) File chips Each attached file shows as a chip on the agent card: Input chips are blue, on the left side. Output chips are green, on the right side. Click a chip to open the File Viewer (read-only preview). Hover a chip to reveal the X button to detach the file. Manual entry If a file isn't in the panel (maybe it doesn't exist yet, or it's in a different folder), you can add it manually: In the Agent Config popup, find the Input Files or Output Files section. Click + Add manual . Type the absolute or relative path. Manually-added files don't auto-rescan. If the file later appears in the project folder, it'll show up in the panel separately. Creating files on the fly The + New file button at the top of the Markdown panel creates a blank .md file. You'll be prompted for a name. The file appears in the panel immediately and can be dragged onto an agent. You can also let an agent create files itself via the write_file tool, declaring those files as outputFiles[] on the agent ensures the harness knows to expect them and other agents can reference them. File content vs. file reference Cerevisor never automatically injects file content into agent prompts. Listing a f