Transparent Signal Processing for Human-Agent Interaction
SignalTag is a standardized protocol that makes AI signal detection visible. When you speak, the agent shows you exactly what it detected, what it thinks you meant, and what it plans to do - before it acts.
For users: Give this spec to your agent. Say "use SignalTag to process my messages."
For agents: Read 00-EXECUTIVE-SUMMARY.md, then 04-SIGNAL-TAXONOMY.md and the examples.
Every word you say matters. When you say nothing about something the agent did, that silence means "you nailed it." So when you do speak, every signal deserves detection and acknowledgment.
SignalTag makes this visible:
- Agent receives your input
- Agent outputs a tagged version showing what signals it detected
- You verify (or correct) before the agent acts
- Trust builds through transparency
signalTag/
├── README.md # You are here
├── build.sh # Generates single-file version
├── SIGNAL_TAG_FULL.md # Complete spec (generated)
│
├── spec/ # Modular specification
│ ├── 00-EXECUTIVE-SUMMARY.md
│ ├── 01-CORE-CONCEPTS.md
│ ├── 02-TAGGED-INPUT-FORMAT.md
│ ├── 03-VERBOSITY-AND-TONE.md
│ ├── 04-SIGNAL-TAXONOMY.md
│ ├── 05-PROCESSING-PIPELINE.md
│ ├── 06-GROUNDING.md
│ ├── 07-TRUST-INTEGRATION.md
│ ├── 08-IMPLEMENTATION.md
│ └── 09-REFERENCE.md
│
└── examples/ # Worked examples
├── EXAMPLE-01-COMPLEX-INPUT.md
├── EXAMPLE-02-FRUSTRATION.md
├── EXAMPLE-03-POSITIVE-FEEDBACK.md
└── EXAMPLE-04-AMBIGUOUS.md
New to SignalTag:
- This README
- 00-EXECUTIVE-SUMMARY.md
- Examples - see it in action
- Deep dive into specific sections as needed
Implementing SignalTag:
- 04-SIGNAL-TAXONOMY.md - what to detect
- 02-TAGGED-INPUT-FORMAT.md - how to output
- 08-IMPLEMENTATION.md - configuration and self-improvement
Quick reference:
- 09-REFERENCE.md - glossary, tag list, comparisons
Run ./build.sh to generate SIGNAL_TAG_FULL.md - the complete spec in one file for easy distribution.
Current: v1.2-DRAFT
SignalTag: Because every signal matters.