Repository navigation
Add "How many agents does the user see?" to agentic product design - #2
Merged
Merged
Conversation
Multi-agent decomposition is an engineering and governance decision. The user should still get one conversation and one resolution — a product that exposes its internal topology as hand-offs has converted an architecture into a user problem, which is the same mistake as exposing the model. Reconciles this with two things already in the file: named persistent agents (the messaging-a-coworker model) are a genuine exception, and attribution is by system rather than by internal boundary — the trace needs per-agent resolution, the surface does not. Checklist gains a topology-leakage item (now 14). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01J5FFKbhpgSWdiBJRaR7Xa8
kalyvask
marked this pull request as ready for review
August 16, 2026 17:10
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The gap
agentic-product-design.mdcovered how an agent should behave, but not the question a PM hits the moment the system becomes multi-agent: does the internal topology surface to the user?What's added
One section, plus a checklist item.
The worked case: "my order arrived damaged, I was charged twice, and I want this resolved." One problem to the customer; internally it may take six capabilities — order and transaction retrieval, duplicate-payment investigation, return policy, resolution choice, execution, verification.
The position: the customer should still see one conversation and one resolution. Internal decomposition is an engineering and governance decision (separate permissions, separate evals, separate release cadences) and is not a mental model the user asked for. A product that exposes its org chart as a hand-off queue — "transferring you to the Billing Agent" — has converted an architecture into a user problem, which is the same mistake as exposing the model.
Two exceptions, both reconciled with material already in the file:
Checklist gains a topology-leakage item (now 14): does the user get one coherent resolution, or does decomposition leak as hand-offs, repeated context, and shifts in voice?
Companion change
The architecture-side treatment of the same material — boundary test, orchestration patterns, MCP vs. A2A, delegation contracts, governance/observability — landed separately as framework 17 in
fde-simulation(PR #2). This repo gets only the user-facing slice, which is the part a PM owns.Checks
All relative links resolve.
Generated by Claude Code