Skip to main content

Agent interaction research

experimentalresearchpublished

A reproducible empirical study of which human turns in agent sessions carry judgment and which can become deterministic or delegated workflows.

Locked revision
a95ea8fefa46
Latest release
Not recorded
No release exists in the current snapshot
Primary outcome
Understand
Audience
Evaluator, Researcher, Developer

Recommended start

Begin with the repository-owned adoption path.

experimentalunderstand

Agent interaction research

Extract a frozen local corpus and reproduce the measurement and workflow-mining stages without model calls.

Path
source build
Time
about 10 min

Needs: Python 3.13, uv, Task

Reproduce the deterministic stages

Audience Evaluator, Researcher, Developer

Public surface

Continue into declared documentation and machine interfaces.

APIs and schemas

No machine specification is declared.

Feeds

No repository feed is declared.

Declared relationships

See what this project points to—and what points here.

Outgoing

  • Informs Agentic PrinciplesGrounds principles in observed agent sessions
  • Informs AEPIdentifies work that benefits from governed workflows
  • Informs Agent PluginsIdentifies repeatable guidance and tool paths