Do this once per research project, with a configured agent. Your supervisor must approve the service and project data it can read. Never share passwords, tokens, or private keys.
1. Find The File#
Reuse your onboarding-IT checkout if the agent can read it. Otherwise, use
the GitHub SSH setup and clone procedure
on the machine where the agent can read project files. Keep this checkout
outside the Passport practice folder. The lab repository requires access;
cloning it is project setup, not a Passport completion requirement.
In its agent_wiki folder, find research_workflow.md and copy the full path,
including the filename. The briefing is the entry point; its links supply the
procedures relevant to your task.
Keep the whole checkout on the machine where the agent runs: it needs the linked procedures too. For Claude Code on Euler, follow step 4 of its setup.
2. Copy This Prompt#
Replace <FULL_PATH> below with the full path to
research_workflow.md you copied in step 1, including the filename.
Open your project and paste the prompt into the agent's chat, not the terminal:
Read "<FULL_PATH>" and follow the linked lab procedures relevant to this project, alongside the project's own instructions.
3. Verify#
Understand before accepting AI output: ask the agent which files it read and check the named paths and relevant rules against the actual files. A claimed reading is not proof by itself. To reuse the reference in future sessions, follow the briefing's instructions for your project's agent setup.
If it cannot, check file access and your tool's instruction settings (Claude Code, VS Code). Still stuck? Send the error message to your supervisor or lab IT, but remove any passwords, tokens, private keys, or confidential information before sending.
Lab context is not permission to act: keep reviewing commands and results.
Sources verified: 2026-10-10. Review by: 2027-01-10. Owner: IDEAL Lab IT.