Skip to content

Add Agent input/output mounts at /agent/input and /agent/output - #18

Open
devzeebo wants to merge 1 commit into
mainfrom
feat/agent-io-mounts
Open

Add Agent input/output mounts at /agent/input and /agent/output#18
devzeebo wants to merge 1 commit into
mainfrom
feat/agent-io-mounts

Conversation

@devzeebo

@devzeebo devzeebo commented Sep 8, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Adds agent.input / agent.output (AgentFs) backed by host temp dirs, bind-mounted on every run at /agent/input (ro) and /agent/output (rw)
  • Clears output at the start of each run(); input persists until clear(); workspace mount unchanged
  • Documents the API and re-exports the new constants/AgentFs from agent-gwt

Test plan

  • pnpm --filter clanker-cleanroom test
  • Write a file via agent.input.write, run with a stub/docker runner, confirm docker args include /agent/input:ro and /agent/output
  • Confirm output is empty after a second run() when nothing new was written by the agent
  • Confirm path escape (../) is rejected by AgentFs

Give each Agent host temp trees to write before run and read after, without changing workspace mounting.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant