Path 4 of 6Practitioner6 lessons · 57 minReviewed 2026-11-10
Securing AI coding assistants in an organisation
What Claude Code, Codex, Copilot, Cursor and Gemini CLI can reach on a developer machine, how their approvals and sandboxes work and fail, and how to enforce policy, MCP allowlists, supply chain controls and telemetry across a fleet.
For: Security, platform and IT teams rolling out AI coding tools; engineering leads
- 01 What a coding agent can reach on a developer machine
Inventory the data, credentials and capabilities a coding agent inherits from the developer's account, including the plaintext transcripts it writes itself.
9 min · 4 questions
- 02 Permissions, approval modes and sandboxing
Compare the approval and sandbox controls in Claude Code, Codex and Cursor, and learn from two real bypasses why approvals alone do not contain an agent.
10 min · 4 questions
- 03 Managed settings and policy across a fleet
Find the layer each coding tool reserves for administrators, deliver a baseline policy to every machine, and plan for the gaps that managed settings do not close.
9 min · 4 questions
- 04 MCP allowlists for developer tools
Choose between a fixed MCP deployment, an approved catalogue and a denylist, write entries that match what actually runs, and close the repository and automation paths that skip approval.
10 min · 4 questions
- 05 Supply chain: malware that drives your agent, poisoned tools and instruction files
Learn from s1ngularity, Shai-Hulud, the Amazon Q extension compromise, package hallucination and hidden-Unicode rules files, and map each to a control you can deploy.
10 min · 4 questions
- 06 Inventory, telemetry and the fleet incident
Know which agents, versions and MCP servers run on which machines, export the telemetry that shows what they did, and rehearse the response when one of them turns out to be compromised.
9 min · 4 questions
Lessons cite primary sources and were last checked on the date shown in each lesson. Corrections are welcome by email (footer). Other paths: Fundamentals · LLM applications · Agents & MCP · Red teaming · Detect & respond