Specialist sub-agents
A specialist sub-agent is the AI Assistant focused on a single systems-engineering role. Each specialist starts from a curated system prompt and a narrower set of tools, so the assistant stops behaving like a generalist and starts behaving like a domain expert — speaking the role's vocabulary and reaching for the right tools first.
How to launch one
Each specialist is launched from a slash command in the chat:
- Open the AI Console (the AI button in the top bar, or Cmd/Ctrl+K) or go to
/ai-assistant. - Type
/in the chat input to open the command palette. - Pick the specialist (for example /requirements-analyst) and add your request.
The assistant stays in that role for the conversation. To return to the generalist, type /clear or start a new chat — specialist context does not leak across conversations. See slash commands for the full list.
When to use a specialist
Use a specialist when you know which systems-engineering role the work belongs to. Use the generalist when you don't. A specialist applies the conventions you would expect from someone in that role and avoids wandering into unrelated tools.
The five specialists
Requirements Analyst
/requirements-analyst
Author and audit requirements: EARS rewrites, decomposition into atomic children, duplicate detection, and completeness checks.
Rewrite REQ-014 in EARS form with a measurable acceptance criterion.
Audit the requirement set for compound, ambiguous, or untestable statements.
Split REQ-027 into atomic children and propose ids.
V&V Engineer
/vv-engineer
Plan and audit verification: allocate verification methods (T/A/I/D), generate test cases, and check the evidence chain and RVTM coverage.
Allocate TAID methods for every unverified critical requirement.
Draft a test case for REQ-031 with its scenario, expected results, and pass/fail criteria.
Audit the RVTM and list requirements with no verification evidence.
Risk Analyst
/risk-analyst
Discover, score, and mitigate risks: surface risks implied by the project, compute residual scores, and draft mitigation plans.
Identify safety-critical risks implied by the current requirement set.
Score residual risk for the top five mitigations and recommend acceptance or further action.
Convert the recent blocker issues into tracked risks with mitigation plans.
Project Controls
/project-controls
Manage schedule health: detect slip, analyze the critical path, and assess resource and milestone health.
Identify tasks at risk of slipping and propose recovery actions.
Explain why each task on the critical path is on it.
Flag milestones whose health is trending red and explain the drivers.
Architect
/architect
Keep architecture coherent: maintain consistency across CONOPS, requirements, and diagrams, and reason about interfaces.
Check that each CONOPS operational scenario references the requirements it depends on.
Reconcile the requirement set against the diagrams that realize it and flag gaps.
Propose an interface allocation table from the current architecture.