Wow, the execution itself worked lightning-fast and the UI is awesome: 

================

Playbooks
Luo uusi LXC-kontti
Palvelin-diagnostiikka
Yritysten yhteystietojen haku
Palvelin-diagnostiikka
Hakee LXC-kontin IP:n Mind Mapista ja suorittaa SSH-diagnostiikan.
Execute Playbook
Initiating execution sequence...
[*] Requesting secure session clone...
[+] Session established: 12d5303286224f25ae918860533ded85
[+] Worker allocated: idiot.c#api-17
[*] Hydrating session context with parameters...
[*] Dispatching playbook [sysadmin_diag] to Omega Lexer...

=== EXECUTION STDOUT ===
[*] AST REWRITER: Compiling Block IFs to GOTO logic (Stack-based)...
>> [REPL] Syote: EXECUTE_PLAYBOOK sysadmin_diag
>>> [HTTP] POST http://127.0.0.1:9500/playbooks/match
[?] Komentoa ei tunnistettu. Yritetaan paatella (LLM Fallback)...
UNKNOWN_INTENT

====================

But the playbook itself needs input/parameters to actually know what to
target?

How can we design it to do that? 

You have actually already solved the parameter problem! Our architecture
handled it perfectly, but the system just needs a tiny tweak to how it
"speaks" to the backend.

How the Parameters Actually Work (Under the Hood)
When you clicked a playbook in the UI, our JavaScript read the "parameters"
dictionary from the Playbook's JSON schema and dynamically rendered the HTML
input fields.

When you clicked Execute, this exact sequence happened:

Capture: The UI scraped all the values you typed into those dynamic form
fields into a payload JSON object.

Hydrate: It sent that payload via PATCH /session/{sessionId}/data. If you look
at your main.py backend, that endpoint directly injects those parameters into
SESSIONS[sess_id]["context"].

Execution Context: When the POST /repl/evaluate request was fired, it passed
the entire context object to the omega-lexer.

The playbook has your parameters! The MESH BASIC engine can read them directly
from the isolated session memory.

Why did it fail with UNKNOWN_INTENT?
Look closely at the MUD output:

>> [REPL] Syote: EXECUTE_PLAYBOOK sysadmin_diag
>>> [HTTP] POST http://127.0.0.1:9500/playbooks/match
[?] Komentoa ei tunnistettu. Yritetaan paatella (LLM Fallback)...
UNKNOWN_INTENT

The Omega Lexer is an NLP (Natural Language Processing) engine. It doesn't
understand hardcoded system commands like EXECUTE_PLAYBOOK. When it saw that
input, it extracted "EXECUTE_PLAYBOOK" as the verb, fired it at your
/playbooks/match API, scored exactly 0 points against the playbook's real
triggers, and gave up!

The Fix: Mimicking Natural Language
We don't need to change the backend. We just need to update our frontend UI to
dynamically read the trigger.verbs and trigger.keywords from the Playbook's
JSON schema, and feed that to the Lexer to guarantee a 100% match score!