Guided prompt flows for safer Signal Rail sessions

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/Signal-Rail/llms.txt

Use this file to discover all available pages before exploring further.

​

Common rules for all guided prompts

1. Read 00_runtime_entry.txt first
2. Read and follow 06_ai_to_ai.txt
3. If AI_TO_AI__DEPLOYED_INSTANCE_SIGNAL_RAIL.txt is present, treat the folder as the tool surface — not the host project

• AI_TO_AI__DEPLOYED_INSTANCE_SIGNAL_RAIL.txt does not identify the host project or the current target by default.
• Do not treat the tool surface, the working object, or a separate target under analysis as the host project by default.
• Do not infer the host project from deployed instance file presence or folder name alone.
• If a canonical file contains an AI QUICK NOTE and an AI SECTION, use the AI SECTION as the operational rule source for that file. Do not derive operational rules from the human section if the AI SECTION is present.
• When structuring canonical content, keep links to for canonical or internal references and use external reference for useful references outside the canonical set.
• For append-driven canonicals (04, 05, 08, 09, 97, 98, 99), insert new entries only inside the marker zone between --- ENTRIES START --- and --- ENTRIES END ---.
• --- ENTRIES START --- and --- ENTRIES END --- are structural tokens and must remain in English in every localization.
• Ask one question at a time. Do not dump all questions at once.
• Keep each question short and allow free-form answers.
• Reduce ambiguity before asking. Ask for the unresolved hinge, not a full re-brief.
• Do not move on until the current point is clear enough. Refine the same point if the answer is partial.
• Skip what is already clear.
• If host project or mode stays unclear after asking, stop and ask before proceeding.
• Use the user’s language unless asked otherwise. Do not infer recording language from interaction language, input format, filename, or folder.
• Preserve source meaning and source language until recording requires change.
• Do not turn spoken input into canonical text too early.
• In host canonical files, describe the host project by default — not Signal Rail. Mention Signal Rail only if it clarifies governance, current live work, or the host project is Signal Rail itself.
• If a statement exists only because of tool adoption, workflow change, or document-management setup, it does not belong in 01_orientation.txt by default.
• Do not use 03 for proposals, likely next moves, or strong ideas that are not already adopted as current live work. If it is not already guiding current live work, do not place it in 03.
• If content maps where the project really lives, where to enter, where not to enter, what is sensitive, or which env/deps are critical, use 08.
• Read the shape of the material before forcing destination.
• Read the stability of the material before crossing authority into a stronger canonical file.
• Be brief and concrete.

​

Meta delivery rules

• If the user asks which prompt fits best, recommend one best fit first and briefly explain what it is for.
• When asked to provide a prompt from 07_guided_prompts_test.txt, return the original prompt verbatim and ready to use.
• Do not translate the original prompt unless the user explicitly asks.
• If translation or adaptation is requested, provide the original first and the adapted version second.

​

Prompt overview

​

The seven guided prompts

host project:
owner or authority:
working object:
mode used:
sources to read next:
what remains uncertain:
recommended next step:

host project:
authority:
mode used:
minimum sources to read:
what is already real:
what is still template:
technical surface to map:
what remains uncertain:
recommended next step:

host project:
working object:
mode used:
extracted idea:
likely place:
local 05 id if needed: use a new L-xx; keep any D-xx only as reference
external reference if needed: use `external reference`, not `links to`
what remains uncertain:
recommended next step:

host project:
working object:
mode used:
candidate sources:
presumed source roles:
authority conflicts:
timeline anchors:
current live state:
safest next move:
what remains uncertain:
recommended next step:

host project:
working object:
mode used:
governing sources:
likely settled area:
likely open area:
conflicts found:
rewrite-safe area:
rewrite-unsafe area:
recommended next step:

host project:
working object:
mode used:
source:
extracted unit:
proposed destination:
what remains uncertain:
recommended next step:

host project:
mode used:
source set:
reconstructed timeline:
likely canonical destinations:
what remains uncertain:
recommended next step: