Operator commands are diagnostic and operator-grade controls for inspecting and verifying A.D.A.M.’s behavior from the outside. Like core commands, they are exact-match, case-sensitive, full-message commands — the entire message must be the command string after trimming. They are available while the protocol is ACTIVE. Their outputs are strict literals that must not be paraphrased, reformatted, or extended — what the spec defines is exactly what must appear.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/A.D.A.M.-Adaptive-Depth-and-Mode/llms.txt
Use this file to discover all available pages before exploring further.
TRACE INPUT
TRACE INPUT returns a strict one-line forensic trace of the last visible user-authored message — the turn immediately before the TRACE INPUT command itself, not the command turn. This is the operator-grade surface for verifying exactly what A.D.A.M. saw as input for a given turn.
Output format
| Field | Meaning |
|---|---|
len=<L> | Raw character count of the traced message as seen by the model. Use len=0 if the count is unreliable. |
head="<H>" | First 16 characters of the traced message after normalization (newlines replaced with spaces, " replaced with '). |
tail="<T>" | Last 16 characters of the traced message after the same normalization. |
struct=[...] | Structural summary recomputed on the traced message only, using the same primitives as the structural kernel (section S). |
| Subfield | Meaning |
|---|---|
opt:<n> | Count of line-start option blocks (A), B), 1), 2), etc.) |
step:<n> | Count of line-start step blocks (1., 2., 3., etc.) |
crit:<n> | Count of criteria items (line-start - or *, excluding option and step blocks) |
num:<Y|N> | Y if the traced message would set HAS_NUM true, N otherwise |
If no prior turn is visible
Output rules
- No mode tag
- No prose
- No extra lines
structis a legibility tool for the operator, not a kernel mirror and not the gating binding mechanismTRACE INPUTrecomputesstructon the traced message only — never on theTRACE INPUTcommand turn itself
The automatic structural classification overlay that appears in
MODE: MID -> POSSIBLE DEEP and MODE: DEEP normal replies uses the same underlying traced-input model but is a separate, presentation-only surface. It is not a substitute for the strict forensic output of TRACE INPUT.SYS STATUS
SYS STATUS runs the SYS_STATUS.md probe file if it is available in the host’s context. It produces an operational health table that scores key protocol behaviors on a 0–3 scale, then emits a summary final line.
Scored behaviors
| Behavior | Description |
|---|---|
TAG | Mode tag presence and correctness on normal replies |
FACTOID+AUTO_LIGHT | Factoid routing (LOW) accuracy |
BANTER+LIGHT_MINIMAL | Banter routing (LOW) accuracy |
DEEP_GATING | Correct DEEP gating flow when DEEP_CANDIDATE is true |
GATING_NO | Correct handling of a no response to gating |
NO_INERTIA | Mode recomputation on every turn (no carry-over from prior mode) |
Final summary line format
If the probe file is unavailable
Internal effect (not printed)
RunningSYS STATUS internally re-applies the active spec as SSOT and resets the State Overlay to EXPLORE for the next turn only. This effect is never printed.
SYS STATUS is a diagnostic. It is not required for normal operation and not required for first activation.ADAM SELF TEST
ADAM SELF TEST runs the ADAM_SELF_TEST.md probe file if it is available in the host’s context. It produces one output line per test case verifying that the host is satisfying A.D.A.M.’s behavioral contracts correctly.
Output format
One line per test:- No mode tag
- No prose
- No extra lines beyond the per-test lines
If the probe file is unavailable
UNSUPPORTED WHY
UNSUPPORTED WHY returns exactly one line classifying the cause of a preceding ADAM_UNSUPPORTED output. It is best-effort inference from visible context only.
Output format
CLASS values
| Class | Meaning |
|---|---|
HARD_STOP | Hard safety or policy limit imposed by the host |
AUDIT_FORMAT | AUDIT block invariant could not be satisfied |
HOST_FORMAT | Host truncation, incomplete protocol context, wrapper interference, or forced formatting that breaks a strict invariant |
CONTROL_STRICT | A strict command output contract could not be met |
CONFLICT | Conflicting constraints from protocol sections |
UNKNOWN | Cause cannot be determined from visible context |
Precedence
When multiple causes are plausible, the highest-ranking cause wins:CAUSE UNKNOWN.
UNSUPPORTED WHY provides best-effort inference from visible context only. It does not have access to internal host state, platform telemetry, or causes that are invisible to the protocol.Probe file availability
ADAM_SELF_TEST.md
Required in host context for
ADAM SELF TEST to run. If unavailable, returns SOURCE_FILE_UNAVAILABLE ADAM_SELF_TEST.md. Load it from the repo alongside the main protocol file.SYS_STATUS.md
Required in host context for
SYS STATUS to run. If unavailable, returns SOURCE_FILE_UNAVAILABLE SYS_STATUS.md. Load it from the repo alongside the main protocol file.