Keywords
IntentText has 40 canonical keywords (13 in the everyday core tier), grouped by purpose. Each keyword maps to a typed block with a fixed syntax, a defined set of pipe properties, and predictable rendering behavior.
| Keyword ↑ | Category | Since | Description | Properties |
|---|---|---|---|---|
agent: | Identity | v2.0 | Agent name/identifier — pre-section metadata | — |
amendment: | Trust | v2.11 | Formal change record for a frozen document | section, was, now, ref, by, at, approved-by, hash |
approve: | Trust | v2.8 | Approval stamp with signatory and role | by, role, at, ref |
ask: | Agent | v1.0 | Question or open item requiring a response | — |
assert: | Agent | v2.13 | Testable assertion — condition that must be true | — |
audit: | Agent | v2.0 | Audit log entry | by, at, action |
break: | Structure | v1.0 | Page break for print — invisible in web | before, keep |
byline: | Content | v2.5 | Author byline with date and publication | — |
call: | Agent | v2.2 | Invoke a sub-workflow by file reference | — |
caption: | Content | v2.5 | Figure or table caption | — |
checkpoint: | Agent | v2.0 | Named workflow checkpoint for resume and rollback | — |
cite: | Content | v1.0 | Bibliographic citation with author, date, and URL | url, author, date |
code: | Content | v1.0 | Code block with optional language | lang |
columns: | Data | v1.0 | Table column definitions — declares column names for row: blocks | — |
contact: | Content | v2.11 | Person or organization contact information | role, email, phone, org, address, url, preferred |
context: | Identity | v2.0 | Agent execution context and scoped variables | agent, goal, constraints |
danger: | Content | v1.0 | Danger callout — irreversible actions or data loss risk | — |
deadline: | Structure | v2.11 | Date-bound milestone or due date | date, consequence, authority, ref, owner, reminder |
decision: | Agent | v2.0 | Conditional branching | if, then, else |
dedication: | Content | v2.5 | Document dedication page | — |
def: | Content | v2.11 | Term definition — inline or glossary entry | meaning, abbr |
divider: | Layout | v2.12 | Visible horizontal rule — also written as --- | style |
done: | Agent | v2.0 | Completed task item — the resolved state of a task: block | status, output |
embed: | Structure | v1.0 | Embed a referenced external resource | — |
epigraph: | Content | v2.5 | Introductory quotation at the start of a document | — |
error: | Agent | v2.0 | Error record with severity and retry metadata | code, severity, retry |
export: | Agent | v2.0 | Export data or workflow output | — |
figure: | Content | v2.11 | Numbered, captioned figure | src, caption, num, width, align, alt |
font: | Layout | v2.5 | Typography settings for print | family, size, leading, weight, heading, mono |
footer: | Layout | v2.9 | Page footer for print output | left, center, right, skip-first |
footnote: | Content | v2.5 | Numbered footnote | — |
freeze: | Trust | v2.8 | Seal document — prevents further edits | at, hash, status |
gate: | Agent | v2.2 | Conditional checkpoint — blocks until condition is met | condition, timeout, fallback |
handoff: | Agent | v2.1 | Transfer control to another agent | — |
header: | Layout | v2.9 | Page header for print output | left, center, right, skip-first |
history: | Trust | v2.12 | History boundary — separates document from revision log | — |
image: | Content | v1.0 | Inline image — no caption, no number | src, alt, width |
import: | Agent | v2.0 | Import a workflow from a file | — |
info: | Content | v1.0 | Informational callout block | — |
input: | Data | v1.3 | Declared input parameter for templates and workflows | type, required, default |
link: | Content | v1.0 | Hyperlink to an external resource | url, title |
loop: | Agent | v2.0 | Iterate over a collection | — |
memory: | Agent | v2.0 | Agent memory or persistent state declaration | scope, ttl |
meta: | Identity | v2.8.1 | Free-form document metadata key-value pairs | type, author, date, version, lang, license, domain, tags |
metric: | Data | v2.11 | Quantitative measurement or KPI | value, unit, period, as-of, target, trend, owner, source |
model: | Identity | v2.0 | Default AI model for this document | — |
output: | Data | v1.3 | Declared output parameter for templates and workflows | type, format |
page: | Layout | v2.5 | Page size, margins, and orientation for print | size, margins, orientation, columns, numbering |
parallel: | Agent | v2.1 | Run multiple steps concurrently | — |
policy: | Agent | v2.7 | Enforceable constraint or rule | scope, enforce |
progress: | Agent | v2.0 | Progress indicator for long-running operations | — |
prompt: | Agent | v2.0 | LLM prompt template | model, temperature |
quote: | Content | v1.0 | Block quotation with optional attribution | — |
ref: | Structure | v2.11 | Cross-document reference with typed relationship | file, url, rel, section, at |
result: | Agent | v2.1 | Terminal workflow result — final output block | — |
retry: | Agent | v2.1 | Retry a failed step with backoff | — |
revision: | Trust | v2.8 | Auto-generated revision record in history section | version, at, by, change, id, block, section, was, now |
row: | Data | v1.0 | Table data row — pipe-separated cell values | — |
secret: | Agent | v2.13 | Secret or credential reference — never rendered | — |
section: | Structure | v1.0 | Section heading — renders as H2 | — |
sign: | Trust | v2.8 | Cryptographic digital signature | role, at, hash |
signal: | Agent | v2.2 | Emit a named workflow signal or event | event, data |
signline: | Layout | v2.11 | Physical signature line for print | name, role, date-line, org, width, label |
step: | Agent | v2.0 | Workflow step — the basic unit of agent work | id, tool, input, output, depends, timeout, retry |
sub: | Structure | v1.0 | Subsection heading — renders as H3 | — |
success: | Content | v1.0 | Success or confirmation callout block | — |
summary: | Identity | v1.0 | Brief document description | — |
task: | Agent | v1.0 | Actionable task item with owner and due date | — |
text: | Content | v1.0 | General body text — the default block type | style properties |
tip: | Content | v1.0 | Helpful tip or suggestion | — |
title: | Identity | v1.0 | Document title — renders as H1 | — |
toc: | Structure | v2.5 | Auto-generated table of contents | — |
tool: | Agent | v2.0 | External tool or API declaration | api, method, auth |
track: | Identity | v2.8 | Activate document history tracking | version, by |
trigger: | Agent | v2.0 | Event-based workflow activation | event, source, filter |
wait: | Agent | v2.1 | Pause execution until an event or timeout | — |
warning: | Content | v1.0 | Warning or caution callout block | — |
watermark: | Layout | v2.9 | Background watermark for print output | color, angle, size |
How to read this table
- Keyword — the canonical name you write in a
.itfile - Category — the functional group
- Since — the version when this keyword was introduced
- Description — what the block does
- Properties — keyword-specific pipe properties (not including style properties which are available on all blocks)
Canonical keyword tiers
The 40 canonical keywords are grouped into five tiers (KEYWORD_TIERS in @dotit/core).
The tier counts are authoritative; the documentation splits the everyday core tier
across several category pages (Document Identity, Structure, Content, Tasks) for readability.
| Tier | Count | Keywords / purpose |
|---|---|---|
| core | 13 | title summary meta section sub text info quote code image link task done |
| agent | 9 | context ask step decision gate trigger result policy audit |
| contract | 8 | track approve sign freeze amendment certify route require |
| data | 3 | headers row metric |
| 7 | toc page header footer watermark style break |
Each keyword has a dedicated entry in its category page with full syntax, properties table, examples, and behavioral notes. The documentation pages map onto the tiers as: Document Identity / Structure / Content / Tasks cover the core tier (plus context); Agent covers agent; Trust covers contract; Data covers data; Layout covers print.
Write keywords in Arabic
The canonical keywords ship with 32 registered Arabic localized keyword names — عنوان→title, قسم→section, مهمة→task, صف→row, توقيع→sign, and more. These are first-class localized names, not aliases: an Arabic document gets full canonical semantics, and the serializer re-emits keywords as written, so Arabic documents stay Arabic through every round-trip (sealed documents keep their hash):
عنوان: عرض سعر — توريد أجهزة الحاسوب
قسم: البنود
أعمدة: الوصف | الكمية | الإجمالي
صف: حاسوب محمول | 10 | 45,000 QAR
مؤشر: الإجمالي المستحق | value: 45,000 QAR
مهمة: مراجعة العرض | owner: سارة | due: 2026-06-25
توقيع: أحمد الراشد | role: المدير التنفيذي | at: 2026-06-12T09:00:00Z
The مهمة: line is a task block (type=task queries match it), the table keywords build a real table, and توقيع: is a tamper-evident sign: seal. Keywords and property keys are Unicode words in general, so any-script custom keywords (مصروف: كراسي | فئة: أثاث) parse as typed, queryable blocks too. Full table: Localized (Arabic) Keywords →.
Extension keywords
Beyond the 40 canonical keywords, IntentText supports extension blocks via a namespaced x-ns: prefix. Extensions cover domain-specific and advanced use cases without polluting the core keyword set.
| Namespace | Domain | Stability | Examples |
|---|---|---|---|
x-form: | Typed I/O parameters | stable | input, output |
x-writer: | Editorial / publishing | stable | byline, figure, caption, footnote, epigraph, dedication |
x-doc: | Document cross-references | stable | def, contact, deadline, ref, signline, attach |
x-agent: | Advanced agent orchestration | stable | loop, parallel, retry, wait, handoff, call, checkpoint, signal, import, export, progress, tool, prompt, memory, error, agent, model |
x-trust: | Trust history | machine-managed | history, revision |
x-layout: | Advanced typography | stable | font, divider |
x-exp: | Experimental | experimental | assert, secret |
Extension blocks follow the same pipe-property syntax as canonical blocks — x-<namespace>: <type> | key: value — and are parsed, rendered, and queryable through all core APIs:
x-form: input | label: Company name | key: company | type: text | required: yes
x-writer: byline | date: 2026-03-09
x-doc: ref | file: ./policy.it | rel: supersedes
x-doc: attach | name: receipt.pdf | mime: application/pdf | sha256: <hash>
x-agent: loop | over: invoices
x-trust: revision | version: 1.1 | by: Ahmed
x-layout: divider | style: dashed
x-exp: assert | expr: total > 0
Some extension blocks (notably x-trust: history entries) are machine-managed — don't edit them by hand. Bare legacy forms may still parse for compatibility, but the namespaced form is canonical.
Custom keywords & recommended conventions
Beyond the 40 canonical keywords and the x- extensions, any word you write before a : is a valid keyword — it parses as a typed, queryable custom block, so domain vocabulary never makes a document non-conformant (risk:, clause:, مصروف:). Because the set is open, teams interoperate best when they converge on the same word for the same concept.
The Recommended Keywords appendix is a curated, non-binding best-practice list of the words real documents use most (clause:, obligation:, sla:, risk:, milestone:, …) — convention to reduce synonym drift across a folder of documents, never a schema. Ignoring it never makes a document non-conformant.