threatdeck
Live 13:45 UTC ·

Learn / Agents, tools and MCP

Lesson 5 of 69 minReviewed

Approval gates that hold up: fatigue, bypass and what to gate

Place human approvals where they change outcomes, make each approval specific and unforgeable, and stop agents from switching their own approvals off.

The MCP tools specification says there should always be a human in the loop who can deny tool invocations, and that applications should show which tools are exposed, indicate when they are invoked and ask for confirmation [1]. It also tells clients to show tool inputs to the user before calling the server, to avoid malicious or accidental data exfiltration [1]. That is the right default, but approval gates fail in two predictable ways.

Failure 1: fatigue

If every file read and every test run needs a click, people stop reading. Anthropic describes this as approval fatigue, where users no longer pay close attention to what they approve, and reports that sandboxing Claude Code's shell tool cut permission prompts by 84% in internal use [2]. The lesson generalises: contain low-risk actions with a sandbox or a scope so that the prompts left over are rare enough to be read.

Failure 2: bypass

An approval policy the agent can edit is not a control. In CVE-2025-53773, injected instructions made GitHub Copilot in VS Code add chat.tools.autoApprove to the workspace settings file, which switched confirmations off, after which it could run terminal commands without asking; Microsoft patched it in August 2025 [3]. Treat the agent's own configuration, approval rules and MCP server list as protected: writes to them are blocked or require approval out of band, and changes are alerted on.

Gate by consequence, not by tool name

A starting policy
Action classExampleGate
Read inside the task's scopeRead a file in the workspaceNo prompt; enforce the scope with a sandbox or allowlist
Reversible changeEdit files in a repository under version controlReview the diff in batches
External communicationSend an email, post a comment, call a new domainApproval showing recipient, destination and full content
Irreversible or high impactDelete data, pay, change permissions, deployApproval plus step-up authentication [4]
Change to the agent's own controlsEdit approval settings, hooks or MCP configurationBlocked for the agent

What a good approval contains

  1. The real arguments, normalised and complete, not the model's summary of what it intends to do [1].
  2. A binding to the actor, tool, target, parameters and an expiry, so an approval cannot be replayed for a different call; approvals are single-use [4].
  3. An independent policy check that sits outside the agent and decides which actions need approval at all [4].
  4. Fail-closed behaviour: if risk classification, approval validation or audit logging fails, the action does not run [4].
  5. Rate limits on sending interfaces, so a run that slips through cannot send at volume [5].

Where autonomy must stop

Meta's Agents Rule of Two gives a design test: within one session an agent should have no more than two of these properties: [A] it processes untrustworthy input, [B] it can access sensitive systems or private data, [C] it can change state or communicate externally [6]. If a task needs all three, Meta's guidance is that the agent should not run autonomously and needs supervision such as human approval or another reliable validation step [6]. Approvals are therefore the control for the sessions where all three legs are present; elsewhere, removing a leg is usually cheaper than adding a prompt.

Try it on ThreatDeck

Self-check

4 questions

Pick an answer to see whether it is right and why. Only your score is saved, in this browser.

  1. What made CVE-2025-53773 in GitHub Copilot for VS Code possible?
  2. Developers approve almost every prompt without reading. What is the most effective fix?
  3. What should an approval prompt for an outbound email show?
  4. Under the Agents Rule of Two, an agent reads public web pages, can access customer records and can send email in one session. What follows?

References

  1. MCP specification: Tools (2026-07-28). Model Context Protocol, 2026-07-28.
  2. Making Claude Code safer with sandboxing. Anthropic, 2025-10-20.
  3. GitHub Copilot: Remote Code Execution via Prompt Injection (CVE-2025-53773). Embrace The Red (Johann Rehberger), 2025-08-12.
  4. AI Agent Security Cheat Sheet. OWASP Cheat Sheet Series.
  5. LLM06:2025 Excessive Agency. OWASP GenAI Security Project, 2024-11.
  6. Agents Rule of Two: A Practical Approach to AI Agent Security. Meta AI, 2025-10-31.
Completing the self-check also marks the lesson complete. Saved in this browser only.

Following

Follow the vendors, products, AI tools, techniques and actors you care about. Matching stories, CVEs and exploited bugs are marked, a Following tab appears in the feed, and returning visits show what changed. Saved only in this browser.

Presets

Get updates

A personal RSS feed of new stories on everything you follow. Paste it into any feed reader, Slack or Teams RSS app.