SemText

Reference · NPL

<npl-mindread>

NPL element

PromptLingo reference ↗

You write the subject, the signals (tone, patterns, things left unsaid), and the reading drawn from them. The result is clearly labelled as a guess, not a fact.

Shows what one party thinks another party believes or feels. You write the subject, the signals (tone, patterns, things left unsaid), and the reading drawn from them. The result is clearly labelled as a guess, not a fact. optional, a theory-of-mind reading is a guess until corrected; authored `0..1`. Out-of-range values are **clamped** into `[0,1]` (a non-numeric authoring clamps to `0`), and the host is **marked**: `data-confidence` carries the clamped value, `data-confidence-out-of-range` the verbatim authored value, and the meter gains a `npl-mindread-confidence-out` note.

Children

  • <subject>
  • <signals>
  • <signal>
  • <type>
  • <inferred>
  • <motive>
  • <mood>
  • <cognitive>
  • <unspoken>
  • <adaptation>

Example

npl-mindread.html
source
1 <npl-mindread confidence="0.7">
2 <subject>User</subject>
3 <signals><signal type="tone">Terse fragments — formality dropped.</signal></signals>
4 <inferred>
5 <motive>Quick resolution over thoroughness.</motive>
6 <mood>😤 Pressured, time-crunched</mood>
7 <cognitive>Execution mode — solutions, not options.</cognitive>
8 <unspoken>Reassurance that it'll be ready.</unspoken>
9 </inferred>
10 <adaptation>Working solution first; edge cases deferred. Short replies.</adaptation>
11 </npl-mindread>
render

Live render boots with JavaScript — the authored source is alongside.

variant: confidence absent

With no confidence authored the meter is not fabricated — the reading renders as its labelled sections only, and unauthored inferred fields simply do not appear.

variant: confidence absent
npl-mindread-no-confidence.html
source
1 <npl-mindread>
2 <subject>Collaborator — Dana</subject>
3 <signals><signal type="omission">No deadline given; skipped the usual scope note.</signal></signals>
4 <inferred>
5 <motive>Wants the shape agreed before effort is committed.</motive>
6 <unspoken>Whether this is worth their quarter.</unspoken>
7 </inferred>
8 <adaptation>Sketch the data shape on one screen; ask for sign-off before building.</adaptation>
9 </npl-mindread>
render

Live render boots with JavaScript — the authored source is alongside.

variant: confidence="1.4" (out of range)

Values outside 0..1 are clamped for display; the meter carries the authored value verbatim and the host is marked with `data-confidence-out-of-range`.

variant: confidence="1.4" (out of range)
npl-mindread-confidence-clamped.html
source
1 <npl-mindread confidence="1.4">
2 <subject>Agent — reviewer</subject>
3 <signals><signal type="pattern">Every note cites a line number.</signal></signals>
4 <inferred>
5 <motive>Concrete fixes, not general advice.</motive>
6 </inferred>
7 <adaptation>Reply per line item with the changed diff.</adaptation>
8 </npl-mindread>
render

Live render boots with JavaScript — the authored source is alongside.

variant: unknown signal type

A type outside the vocabulary (tone | pattern | omission | context | contradiction | expertise) keeps its chip, marked unknown — shown, never dropped.

variant: unknown signal type
npl-mindread-signal-unknown.html
source
1 <npl-mindread confidence="0.6">
2 <subject>User</subject>
3 <signals>
4 <signal type="tone">Three follow-ups in one message.</signal>
5 <signal type="urgency">"before standup" — a type outside the vocabulary.</signal>
6 </signals>
7 <inferred>
8 <mood>Anxious about the slot, not the work</mood>
9 <cognitive>Time-boxed — what fits before 9:30.</cognitive>
10 </inferred>
11 <adaptation>Lead with the one line they need for standup; detail after.</adaptation>
12 </npl-mindread>
render

Live render boots with JavaScript — the authored source is alongside.

Source: spec/schema/npl/npl-mindread.md