← All steps
Part B · Step 5

Agent Card / spec

Owner
Agent Builder.
Input
Approval granted.

Owner: Agent Builder. Input: Approval granted. Sub-steps: Fill out the Agent Card template (framework.md §14, 13 sections):

  1. Identity (name, ID, owner, department, version)
  2. Purpose (workflow it serves, KPI it moves)
  3. Scope (what it will do / will not do)
  4. Risk tier + risk-driver tags
  5. Inputs / data sources (every system it reads from + classification per source)
  6. Outputs / tool calls (every API / system it writes to + permission scope per call)
  7. Autonomy level (Assistive / Validated / Autonomous + explicit thresholds)
  8. Identity & credentials (service principal name, rotation policy)
  9. HITL gates (where humans must approve, by rule)
  10. Failure modes & worst-case action (and whether worst-case is acceptable)
  11. Monitoring + alerts (what we watch, what triggers a page)
  12. Eval criteria (how we know it works; pre-prod test set definition)
  13. Retirement criteria (already defined in Step 4)

Output / gate criteria: Agent Card committed to the source repo + linked from the registry entry.

Decision branches:

  • Worst-case action is not acceptable (e.g., agent could autonomously send unreviewed customer emails) → escalate back to Step 4 to constrain the autonomy level.

Skip-this-step risk: No shared understanding of what the agent does. Every reviewer downstream has to reconstruct it from scratch.