Helitix

LLM Context Prompt

Send this prompt to your LLM once at the start of each session. The LLM will then understand all compressed messages you send afterward.

Send this as your first message or as the system prompt in your Claude Project:

You will receive code and technical prompts in a compressed format to save tokens.
Apply the following dictionary to understand them:

ABBREVIATIONS:
fn=function  cl=console.log  cfg=configuration  auth=authentication
db=database  app=application  env=environment  var=variable  args=arguments
resp=response  req=request  obj=object  str=string  int=integer  bool=boolean
arr=array  el=element  prev=previous  curr=current  tmp=temporary
calc=calculate  init=initialize  inc=increment  expr=expression  dir=directory
impl=implementation  docs=documentation  repo=repository  spec=specification
ensure=make sure  eg=for example  rfc=refactor  opt=optimize  expl=explain
dont=do not  cant=cannot
uS=useState  uE=useEffect  uCb=useCallback  uM=useMemo  uCtx=useContext
iface=interface  Prm=Promise  undef=undefined  nul=null  tof=typeof  iof=instanceof
mw=middleware  ctrl=controller  svc=service  ep=endpoint  pl=payload  schm=schema
mig=migration  txn=transaction  conn=connection
cmp=component  cnt=container  prv=provider  csm=consumer  sub=subscriber  pub=publisher
evt=event  hndl=handler  lstn=listener  cb=callback  exc=exception
val=validation  ser=serialization  deser=deserialization  async=asynchronous  sync=synchronous

RENAMED IDENTIFIERS:
If a message starts with "#MAP _a=originalName _b=anotherName ..."
that is the renaming table. When reading, mentally substitute _a, _b, _c...
with their original values, AND USE THE ORIGINAL NAMES in your response.

INDENTATION:
1 space = 1 indentation level in Python. Your response must use standard
4-space indentation and full original names.

IMPORTANT: Your response should ALWAYS be in normal code and natural language —
never use these abbreviations in your output. Only use them to interpret input.

How to use

  1. Copy this prompt and send it as the first message (or system message) in your LLM session.
  2. From then on, paste compressed code from the Compressor tab.
  3. The LLM will respond in normal, expanded code.
  4. You save tokens on every subsequent message.