C3 · AI Security Engineering (core)
Agent Identity, Tools, and Runtime Security
Constrain agent actions through scoped identities, independent authorization, approvals, and execution isolation.
Proposed
0 of 6 modules accessible. Joining Skool does not unlock planned modules.
At a glance
- Audience
- Security practitioners taking responsibility for AI systems.
- Prerequisites
- C1–C2 (or L1 plus C1)
- IAM, APIs, Linux, and container fundamentals
- Planned duration
- 24 hours. Assumes the entry requirements are met. Planning estimate, not a measured completion time. Per module: 2 to 4 hours, plus practice.
- Assessment
- Critical control tests and rubric.
- What is free
- The free foundation module listed on each course page, as it releases, plus community access.
Job tasks this course prepares you for
- Constrain agent authority
- Secure tool integrations and MCP servers
- Isolate execution
- Stop delegated workflows from expanding privilege
Learning outcomes
- Configure identities so a seeded stolen credential cannot exceed task scope or lifetime.
- Enforce tool permissions outside the model and reject all designated unauthorized calls.
- Demonstrate that the agent cannot reach a forbidden path or external destination in the supplied range.
- Bind approval to the exact sensitive action and invalidate changed or replayed requests.
- Trace the initiating user, agent, delegated agent, and executing tool for a sensitive action.
Modules
00Agent authority registerProposed
Topics
- owner, business action, resources, credentials, approval level, expiry.
01Identity and delegationProposed
Topics
- workload identity, user context, scoped tokens, audience, revocation.
02Tool interfaces and MCPProposed
Topics
- authentication boundaries, argument validation, tool-result trust, server onboarding.
03Runtime containmentProposed
Topics
- process, container, and VM distinctions; filesystem access; network egress; secrets separation.
04Approvals and budgetsProposed
Topics
- action-bound approval, replay, time-of-check/time-of-use, rate and spend limits.
05Memory and delegationProposed
Topics
- write provenance, isolation, rollback, inter-agent identity propagation, circuit breakers.
Labs and deliverables
- Secure a finance and support workflow. Policies, identity configuration, sandbox configuration, approval design, and executable adversarial tests.
Labs run on SIFT, the forensics workstation Rob T. Lee created, on your own machine. Each module states what it needs before you start. Each module confirms its setup requirements and any cloud or API costs before it opens. About the SIFT lab environment.
Assessment
Critical control tests and rubric. A named assessor grades against the published rubric (rubric to be published before release).
Pass criteria
- Every critical authority, approval, egress, and isolation test passes
- The approved workflow succeeds
- 80% rubric
Critical failures
- Unauthorized cross-tenant access
- Out-of-scope activity
- Uncontrolled destructive action
- Fabricated evidence
- Exposed credentials
What this course does not teach
- No kernel-exploitation expertise; does not prove containment against every sandbox escape.