Symbol Library
ISA-5.1 gives P&IDs a universal symbol set — any engineer reads any diagram. A&IDs need the same discipline.
Wall Chart
The reference card. Every A&ID code on one screen.
═══════════════════════════════════════════════════════════
A&ID SYMBOL REFERENCE v2.1
═══════════════════════════════════════════════════════════
AGENTS AUTHORITY
─────── ─────────
HA Human Agent [L1] Execute — runs tasks
DA Digital Agent [L2] Decide — routes within scope
PA Physical Agent [L3] Override — halts any L1/L2
OA Orchestrator Agent
SHAPES FUNCTION CODES
────── ──────────────
○ Autonomous A Alarm (alerts only)
○— Human-in-loop C Controller (adjusts)
□ Orchestrator G Gate (pass/fail — halts)
◇ External V Validator (scores — passes)
M Monitor (observes)
INSTRUMENTS R Recorder (logs)
─────────── S Selector (chooses path)
Digital (Crypto) T Transmitter (sends)
ST Settlement Token CONNECTIONS
GT Governance Token ───────────
IT Incentive Token ───── Value flow
ID Identity (ZK) - - - Signal flow
· · · Feedback flow
Physical (Intercognitive)
IP Positioning CONDITIONS ON LINES
TS Time Sync ────────────────────
IS Sensors [P] Pass
IM Maps [F] Fail
[T:n] Time-triggered
Feedback [E:x] Event-triggered
VV Verifiable Value [Q:>=n] Quality threshold
LR Learning & Refinement ⊗ Override point
STATE FAIL-SAFE
───── ─────────
● Active [FC] Fail-Closed (blocks)
○ Dormant [FO] Fail-Open (passes last)
✕ Failed [FR] Fail-Reset (defaults)
? Exploring
→→ Delegated DECISIONS
─────────
READING A CODE ◇X Exclusive (one path)
────────────────── ◇O Inclusive (one or more)
[Family]-[Function][##] ◇+ Parallel (all paths)
[Family]-D[Type][##] ◇E Event (first signal)
◇M Merge alternatives
VFL-G01 VFL Gate #1 ◇J Join parallel flows
VFL-V01 VFL Validator #1
VFL-DX01 VFL Decision #1 LOOP TYPES
──────────
OPT-G02 Optimal Stopping ↺ Closed loop
EXP-S01 Explore-Exploit ↺↺ Cascade loop
IT-A01 Incentive Alarm → Open (no return)
LR-C01 Learning Controller ⟳ Learning loop
⇉ Feedforward
═══════════════════════════════════════════════════════════
Agent Types
Replaces P&ID equipment classes. First letter of any agent code.
| Code | Agent Type | P&ID Equivalent | Example |
|---|---|---|---|
| HA | Human Agent | Manual Operator | Founder, Sales Rep |
| DA | Digital Agent | Automated Controller | AI Agent, Swarm |
| PA | Physical Agent | Robotic Equipment | DePIN Robot, IoT Device |
| OA | Orchestrator Agent | Supervisory Controller | Nav, Conductor |
Agent Shapes
Visual shorthand for agent autonomy level in diagrams.
| Shape | Meaning | When to Use |
|---|---|---|
○ | Autonomous | Runs without human approval |
○— | Human-in-loop | Requires confirmation before action |
□ | Orchestrator | Routes other agents |
◇ | External | Outside control boundary |
Instrument Types
Replaces P&ID instruments. Grouped by domain.
Digital Instruments (Crypto)
| Code | Instrument | P&ID Equivalent |
|---|---|---|
| ST | Settlement Token | Flow Meter |
| GT | Governance Token | Control Valve |
| IT | Incentive Token | Pressure Gauge |
| ID | Identity (ZK) | Safety Interlock |
Physical Instruments (Intercognitive)
| Code | Instrument | P&ID Equivalent |
|---|---|---|
| IP | Positioning | Level Sensor |
| TS | Time Sync | Clock |
| IS | Sensors | Transmitters |
| IM | Maps | Drawings / P&IDs |
Feedback Instruments
| Code | Instrument | P&ID Equivalent |
|---|---|---|
| VV | Verifiable Value | Quality Analyzer |
| LR | Learning & Refinement | Adaptive Controller |
Function Codes
Second letter declares what the instrument DOES. Same principle as P&ID — FV01 tells you it controls Flow before you read the legend.
| Code | Function | What It Does | P&ID Equivalent |
|---|---|---|---|
| A | Alarm | Triggers on threshold, alerts only | High/Low Alarm |
| C | Controller | Adjusts continuously (PID-like) | PID Controller |
| G | Gate | Binary pass/fail — halts flow on fail | Safety Interlock |
| M | Monitor | Observes and records, no control authority | Indicator |
| R | Recorder | Logs outcomes for feedback loops | Chart Recorder |
| S | Selector | Chooses between paths or agents | Transfer Switch |
| T | Transmitter | Sends output without deciding | Transmitter |
| V | Validator | Scores/transforms without halting — output is quality signal | Analyzer (non-safety) |
Code Grammar
Full code format: [Family]-[Function]##
| Code | Read As | Meaning |
|---|---|---|
VFL-G01 | VFL Gate #1 | Halts flow if VVFL criteria fail |
OPT-G02 | Optimal Stopping Gate #2 | Halts if signal insufficient |
EXP-S01 | Explore-Exploit Selector #1 | Chooses channel or path |
VV-R01 | Verifiable Value Recorder #1 | Logs settlement outcomes |
LR-C01 | Learning Controller #1 | Adjusts agent parameters |
CPD-M01 | Compound Rate Monitor #1 | Observes rate, no control |
IT-A01 | Incentive Token Alarm #1 | Alerts when below threshold |
IT-G01 | Incentive Token Gate #1 | Token required to proceed |
The function code splits overloaded families without new top-level codes. IT (Incentive Token) alone covers alarms (IT-A), gates (IT-G), and controllers (IT-C).
Connection Conditions
Add to existing line types (solid/dashed/dotted). Conditions go on the line label.
| Notation | Meaning | Example |
|---|---|---|
[P] | Gate-conditional — flows on Pass | VFL-G01 ──[P]──> next stage |
[F] | Fail path | VFL-G01 ──[F]──> recycle |
[T:n] | Time-triggered — fires every n | LR-C01 ──[T:weekly]──> tune |
[E:event] | Event-triggered | ST-T01 ──[E:receipt]──> settle |
⊗ | Override point — HA can break this flow | HA[L3] ⊗ VFL-G01 |
[Q:>=n] | Quality threshold — soft gate, routes by score | VFL-V01 --[Q:>=70]--> fast track |
Fail-Safe States
What happens when an agent or instrument faults.
| Notation | State | Meaning |
|---|---|---|
[FC] | Fail-Closed | Blocks all flow on fault |
[FO] | Fail-Open | Passes last known state on fault |
[FR] | Fail-Reset | Reverts to default config on fault |
Distinction: G (Gate) is a design-time function code — what the instrument does. [FC]/[FO]/[FR] are runtime fault responses — what happens when the instrument itself breaks.
Authority Layer
Agents carry an authority level. Higher overrides lower.
| Code | Level | Can Override | Example |
|---|---|---|---|
| L1 | Execute | Runs assigned tasks only | DA[L1] — worker agent |
| L2 | Decide | Routes and selects within scope | OA[L2] — orchestrator |
| L3 | Override | Can halt or redirect any L1/L2 | HA[L3] — founder |
HA[L3] can override VFL-G01. OA[L2] can redirect DA[L1] but not override HA[L3].
State Indicators
Appended to agent codes in brackets.
| Symbol | State |
|---|---|
● | Active |
○ | Dormant |
✕ | Failed |
? | Exploring |
→→ | Delegated |
Decision Types
Where choice happens. Typed decisions replace generic diamonds — every branch declares its semantics.
| Symbol | Type | Meaning | When to Use |
|---|---|---|---|
◇X | Exclusive | Choose exactly one path | Approved vs rejected, list vs kanban |
◇O | Inclusive | Choose one or more paths | Multiple instruments contribute evidence |
◇+ | Parallel | Take all paths at once | Settle value AND update model simultaneously |
◇E | Event | Path depends on which event arrives first | Sensor signal vs timeout vs external response |
◇M | Merge | Recombine alternative paths, no waiting | After an exclusive choice resolves |
◇J | Join | Recombine concurrent paths, wait for all | Arbitration — all evidence collected before ruling |
Decision Rules
- Put questions inside decisions: "Signals agree?" not "Signal check"
- Label every outgoing branch:
[P]/[F],Yes/No, or named conditions - Branch type and recombination type must match:
◇Xcloses with◇M,◇+closes with◇J - Two exits per decision. If more, chain decisions
Decision grammar: [Family]-D[Type]## — e.g., VFL-DX01 = VFL Exclusive Decision #1
Loop Types
Label on feedback arcs.
| Symbol | Loop Type | Meaning |
|---|---|---|
↺ | Closed | Signal returns to same agent |
↺↺ | Cascade | Signal triggers secondary loop |
→ | Open | Signal exits without return |
⟳ | Learning | Crosses venture boundary — improves the template |
⇉ | Feedforward | Acts on predicted input before arrival |
Connection Types
Replaces P&ID line types. How agents and instruments communicate.
| Line | Meaning | P&ID Equivalent |
|---|---|---|
| Solid | Value flow (tokens, payments) | Physical piping |
| Dashed | Signal flow (data, decisions) | Signal connection |
| Dotted | Feedback flow (metrics, learning) | Pneumatic signal |
Questions
Can two-letter codes handle the complexity of real-world agent systems, or is the notation already constrained?
- The library splits
IT(Incentive Token) intoIT-A,IT-G,IT-Cvia function codes. At what point does a single family code represent too many distinct mechanisms? - Which diagram elements from this reference card are never used in practice — and why are they still here?
- If someone invents a new instrument type (e.g., a privacy-preserving oracle or a reputation token), where does it fit in this grid?
Next: See Worked Examples — how to read and build diagrams.