The emsemioverse has invested in training the agent to interpret prompts — the encoding loop, prompt routing, the interpret-message skill, and the signal classification taxonomy. But the interaction is two-sided: the user’s ability to prompt effectively is as important as the agent’s ability to interpret.
The asymmetry
The current architecture optimizes the receiver but not the sender. The agent has a dispatch algorithm, trigger phrases, and fallback logic. The user has nothing — no guide to what makes a good prompt for this system, no feedback on whether their prompt was efficiently interpreted, no way to learn the system’s vocabulary for their own benefit.
This asymmetry means emsenn must learn the system’s affordances through trial and error, which is slow and produces the corrections that the agent then encodes. A prompting guide would let emsenn learn the system’s patterns deliberately, reducing the correction cycle.
What a prompting guide would cover
- Signal classification markers — how to make it obvious whether a message is a command, correction, meta-commentary, or question, so the agent classifies correctly on first pass.
- Skill trigger vocabulary — the phrases that reliably trigger
specific skills, so emsenn can invoke skills without needing to
know the
/slash-commandnames. - Session initialization patterns — how to write a first message that sets context efficiently: what to include, what to omit, how to structure goals.
- Scope signals — how to indicate the size of a task (quick fix vs. session-long project vs. multi-session arc) so the agent allocates attention appropriately.
- Correction patterns — how to correct the agent in ways that produce persistent fixes rather than session-local adjustments.
Relationship to the encoding loop
The prompting guide is itself an encoding: it takes tacit knowledge about how to interact with the system and makes it explicit. This is the encoding loop applied to the interaction protocol itself — the meta-level of the system.
Source
emsenn identified this gap on 2026-03-07: “we have not done much to train me in how to prompt you to facilitate our goals.”