Skip to main content

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.

A mode in Signal Rail is a permission boundary, not a descriptive label. It defines what the agent is allowed to do during the current session. Only one mode may be active at a time. When a mode changes, what is allowed changes with it. If a closed gate reopens during the task — a write is attempted from a read-only mode, or a structure-level touch is initiated without structure mode — stop treating execution as valid and re-enter through the recovery path before continuing.

Mode Reference

ModeWhat it allowsWhat it does NOT allow
orientationReading, minimum read completion, source comparison, uncertainty declaration, frame closure, task clarificationWriting, promotion, structural touches
referenceReading specific canonical files in scopeWriting or rewriting
reviewComparing material, identifying conflicts, surfacing risksWriting, canonical updates
ingestExtracting and classifying units from sourcesWriting to 01, 02, or 06; promotion without destination closure
workExtracting, classifying, proposing destinations, writing to clear destinationsTouching 01, 02, or 06 without structure mode
structureTouching 01_orientation.txt, 02_protocol_freeze.txt, or 06_ai_to_ai.txtMust be explicitly opened; crossing from another mode requires stop-and-ask

Default Startup State

Every session starts in orientation mode. Orientation is read-only. Reading does not authorize action. Understanding does not authorize writing. The agent may complete the minimum read, compare sources, declare uncertainty, close the working frame, and clarify the task — nothing more until the mode is explicitly moved.

Structure Mode

Structure mode is required before touching 01_orientation.txt, 02_protocol_freeze.txt, or 06_ai_to_ai.txt. If a task starts in work mode and a structure touch becomes necessary, stop and ask before crossing. Before entering structure mode, reread 06_ai_to_ai.txt.
Structure mode must be explicitly opened. It does not carry over from a previous session or inherit from work mode. Crossing into structure mode mid-task requires a deliberate stop-and-ask, not an internal mode shift. The agent may not assume structure mode is valid because the task seems to require it.

The Move Order

The operating procedure in 06_ai_to_ai.txt defines a fixed sequence. Steps must be followed in order. Later steps do not authorize skipping earlier ones.
1

Activate

Read 06_ai_to_ai.txt to enable the Signal Rail protocol.
2

Close object

Identify and close the current working object before any mode or container decision.
3

Close mode

Identify and close the active mode for the session. This sets permission boundaries.
4

Minimum read

Complete the minimum read before substantive action: 01_orientation.txt, 03_master_working.txt, and any conditional files triggered by the task.
5

Identify source authority

When multiple sources are in scope, identify which carries governing authority for the touched level.
6

Clarify task

Before extracting or writing, confirm what the task actually requires. Do not infer scope from task volume.
7

Extract

Read source material in the current session. Extract units. Do not route during extraction.
8

Classify

Classify each extracted unit by nature, stability, and level. Classify units, not files.
9

Propose destination

For each classified unit, propose a canonical destination. Do not promote until the destination is clear enough.
10

Write only if clear

Write only when destination, level, and authority are all sufficiently closed. If any remain unclear, hold the unit in 05_latent_ideas.txt.

Operations Vocabulary

The following named operations come directly from 06_ai_to_ai.txt. Each carries a specific meaning in the protocol.
OperationMeaning
read 06_ai_to_ai.txtActivate the protocol; does not authorize bootstrap or writing
preserveSave without losing the right context
read and reworkRead the material and absorb it into the right canonical container
absorbMove content from latent or scattered into one of the live canonical files
parkMove useful but inactive material to 98_parking.txt
archiveMove historical or closed material to 99_archive.txt
unearthReview entries in 05_latent_ideas.txt and choose: absorb, park, or leave alive

Stop-and-Ask Conditions

The following conditions all require stopping immediately and asking for clarification before taking any further action:
  • Host project is unclear
  • Current working object is unclear
  • Touched level is unclear
  • Mode is unclear
  • Authority is unclear
  • Scope is unclear
  • Live source is unclear
  • The append-driven entries contract is unclear on a touched canonical file
  • Classification could change meaning or continuity
  • Promotion could change structure, authority, or current live state
  • Canonicals conflict with each other
  • Canonicals conflict with implementation
  • More than one valid reading remains after review

Mode Boundary Violation

If a protocol rule is violated — for example, writing from orientation mode, or touching a structure-level file from work mode — stop processing immediately and return only: violated rule / blocked action / missing clarification. Do not attempt recovery by continuing with reduced scope.

Build docs developers (and LLMs) love