markform 0.1.23 → 0.1.25

package/examples/markform-demo-playbook.md

@@ -0,0 +1,342 @@
+# Markform End-to-End Example: S&P 500 Company Research
+
+## About This Playbook
+
+### Purpose
+
+This playbook walks an agent through a complete Markform workflow to demonstrate its
+features: from a research idea to a validated, filled, exported structured report with
+typed JSON/YAML data and a JSON Schema.
+
+### Playbook Format
+
+**If you’re a human:** You can read this to get a sense for how you and agents can use
+Markform.
+
+**If you’re an agent:** Follow this playbook step by step.
+
+This playbook works in two ways:
+
+1. **Agent instructions:** Steps an agent can follow start to finish.
+2. **User narration:** “Tell the user” blocks keep the user informed at each phase.
+
+Notes:
+- An agent must have access to use `markform` as a CLI.
+- Commands give output and/or errors.
+  For example, every `markform set` command should output `Form updated: <path>`.
+- An agent or the user can review the result.
+
+### Playbook Scope
+
+**Scenario:** Research an S&P 500 company (default: Apple Inc.). The user provides a
+company name; the agent designs a form from scratch, fills it with researched data,
+validates, and exports.
+
+This playbook has the agent *manually* fill each field the granular `next` → `set` loop,
+researching data along the way.
+This exercises all CLI operations and `set` capabilities.
+
+Markform also supports fully automated filling via `markform fill`, where an LLM drives
+the `next` → `set` loop autonomously.
+That’s out of scope here — this playbook is about the agent researching and filling each
+field itself. Run `markform examples` to see automated filling in action.
+
+## Phase 0: Prerequisites
+
+Before proceeding, confirm you have what’s needed:
+
+1. **Shell access** to run `npm` and `markform` CLI commands
+2. **Web search** or other research tools to look up company data
+3. **File read/write** access to create and edit form files in `/tmp/markform-qa/`
+
+If anything is missing, tell the user what’s needed before continuing.
+
+### Choose the Company
+
+If the user has already specified a company, skip to Phase 1.
+
+Otherwise:
+
+> **Tell the user:** "I have a playbook that walks through Markform, a CLI for
+> structured, validated forms.
+> I’ll design a research form for an S&P 500 company, fill it with real data, validate
+> it, and export the results.
+> 
+> Is there a company you’d like me to research?
+> If you aren’t sure, I’ll just pick Apple (AAPL) as an example."
+
+Wait for the user’s response.
+Use their chosen company (default: Apple Inc.).
+
+## Phase 1: Install Markform
+
+> **Tell the user:** “I’ll install (or update) Markform: a CLI for structured, validated
+> forms that agents fill via tool calls.”
+
+```bash
+npm install -g markform@latest
+markform --version
+```
+
+> **Tell the user:** “Markform [version] is installed.”
+
+## Phase 2: Define the Research Scope
+
+> **Tell the user:** “We’ll build a structured research form for an S&P 500 company.
+> Unlike free-form notes, a form ensures revenue is a number, the ticker matches a
+> pattern, segments are a typed table, and URLs are validated: all enforced
+> automatically. Let me outline what to capture.”
+
+Draft a plain markdown outline of research questions covering at least these areas:
+
+1. **Identification:** Company name, ticker, GICS sector, headquarters, founding year,
+   website
+2. **Financials:** Fiscal year, revenue, net income, market cap, employee count
+3. **Business segments:** Revenue breakdown by segment with percentages
+4. **Leadership:** Key executives with name, title, year appointed
+5. **Competitive position:** Competitors, competitive moats, market position summary
+6. **Sources & verification:** Source URLs, research date, verification checklist, notes
+7. **Source provenance:** A comprehensive table of all sources consulted during
+   research, with four columns: source name, what information it was used for, notes on
+   source quality or reliability, and URL. This turns the form into a verifiable audit
+   trail — every fact is traceable back to its origin.
+
+Save as `/tmp/markform-qa/research-outline.md`.
+
+> **Tell the user:** "I’ve saved the research outline to
+> `/tmp/markform-qa/research-outline.md`. Next I’ll convert it into a Markform: a
+> structured form with typed, validated fields."
+
+## Phase 3: Build the Markform
+
+### Step 3.1: Learn the Syntax
+
+> **Tell the user:** “I’ll read the Markform reference to learn how to turn this outline
+> into a validated form.”
+
+```bash
+markform docs
+```
+
+Read the full output.
+It covers file structure, all 11 field kinds, validation rules, CLI commands, and
+includes a complete example form.
+
+### Step 3.2: Study an Example
+
+```bash
+markform examples --name simple --forms-dir /tmp/markform-qa/ref
+```
+
+Read the example form to see how field kinds, groups, and validation look in practice.
+
+> **Tip:** Run `markform examples --list` to see all bundled examples.
+> You can copy any example with `markform examples --name <id>` for additional
+> reference.
+
+### Step 3.3: Convert to Markform
+
+> **Tell the user:** “I’ll convert the outline into a Markform using the field kinds and
+> validation rules described in the docs.”
+
+Convert the research outline into `/tmp/markform-qa/sp500-research.form.md`. Use the
+“Form Design Guide” and “Best Practices” sections from `markform docs` as reference for
+choosing field kinds, adding validation, and structuring the form.
+
+The last field in the form should be a source provenance table — a comprehensive list of
+sources with four columns: source name, information used, notes on quality, and URL. Use
+a `url`-typed column for the URL so it gets validated.
+This table is `required` — no research report is complete without attribution.
+
+### Step 3.4: Validate and Fix
+
+> **Tell the user:** “I’ll validate the form structure and fix any issues.”
+
+```bash
+markform validate /tmp/markform-qa/sp500-research.form.md
+```
+
+If issues appear beyond `required_missing` (expected: the form is empty), fix the
+`.form.md` and re-validate.
+Repeat until clean.
+
+```bash
+markform inspect /tmp/markform-qa/sp500-research.form.md
+```
+
+> **Tell the user:** “Form structure is valid: [N] fields across [N] groups, ready to
+> fill.”
+
+## Phase 4: Fill the Form
+
+The core workflow: run `next` to see what to fill, use `set` to fill it, repeat.
+
+### Step 4.1: Set the User Input
+
+> **Tell the user:** "The company name field has `role='user'`: it’s the one
+> human-provided input.
+> I’ll set it to start the research."
+
+```bash
+markform set /tmp/markform-qa/sp500-research.form.md <company_name_id> "Apple Inc."
+markform status /tmp/markform-qa/sp500-research.form.md
+```
+
+### Step 4.2: The next → set Loop
+
+> **Tell the user:** "`next` is the field advisor: it prioritizes what to fill and shows
+> `set` examples. I’ll follow its advice, research the data, and repeat."
+
+```bash
+markform next /tmp/markform-qa/sp500-research.form.md
+```
+
+Research accurate data and fill fields using `set`. Continue the `next` → `set` loop
+until the form is complete.
+
+As you fill each group of fields, also populate the source provenance table: append a
+row for each source you consulted, recording what information it provided and its URL.
+This ensures every fact in the form is traceable.
+
+**Use each of these operations at least once during filling to show the full range of
+`set` capabilities:**
+
+| Operation | Example |
+| --- | --- |
+| Single set | `markform set <file> <id> "value"` |
+| Batch set | `markform set <file> --values '{"a": 1, "b": "text"}'` |
+| Table append | `markform set <file> <tableId> --append '{"col": "val"}'` |
+| Table batch set | `markform set <file> <tableId> '[{...}, {...}]'` |
+| Table delete + re-append | `--delete <index>` then `--append` replacement |
+| Validation error | Set an invalid value (e.g., lowercase ticker): verify it's stored but flagged invalid |
+| String list | `markform set <file> <id> '["a", "b", "c"]'` |
+| Multi-select | `markform set <file> <id> '["opt1", "opt2"]'` |
+| Explicit checkboxes | `markform set <file> <id> '{"item1": "yes", "item2": "no"}'` |
+| Skip optional field | `markform set <file> <id> --skip --reason "..."` |
+
+> **Tell the user:** After the validation error: "`set` stores the value but `validate`
+> flags it as invalid.
+> Some constraints (like pattern matching) are checked eagerly by `validate`; semantic
+> issues may surface later.
+> Fix the value and confirm with `validate`."
+
+### Step 4.3: Confirm Completion
+
+```bash
+markform next /tmp/markform-qa/sp500-research.form.md
+```
+
+**Checkpoint:** Should report form complete, 0 required fields remaining.
+
+## Phase 5: Validate and Export
+
+> **Tell the user:** “Form is complete.
+> Let me validate and export.”
+
+```bash
+markform validate /tmp/markform-qa/sp500-research.form.md
+```
+
+**Checkpoint:** Form State `complete`, `invalid=0`, no issues.
+
+```bash
+markform dump /tmp/markform-qa/sp500-research.form.md
+markform export /tmp/markform-qa/sp500-research.form.md --format=markdown
+markform report /tmp/markform-qa/sp500-research.form.md
+markform export /tmp/markform-qa/sp500-research.form.md --format=json
+markform export /tmp/markform-qa/sp500-research.form.md --format=yaml
+markform schema /tmp/markform-qa/sp500-research.form.md --pure
+```
+
+> **Tell the user:** "Exported as:
+> - **Markdown (export):** full rendered form including field instructions — useful as a
+>   standalone document or for review
+> - **Markdown (report):** clean results only, no instructions — useful as a deliverable
+> - **JSON/YAML:** typed structured data for databases or APIs
+> - **JSON Schema:** form structure for code generation or external validation
+> 
+> Every fact in the report is traceable: the source provenance table records where each
+> piece of information came from, with validated URLs.
+> The form doesn’t just add structure — it adds reliability."
+
+## Phase 6: Advanced Operations
+
+### Step 6.1: Clear and Re-fill
+
+> **Tell the user:** “Any field can be cleared and re-filled for corrections.”
+
+Clear a number field, verify it shows as unanswered, then set a new value.
+
+### Step 6.2: List Append and Delete
+
+> **Tell the user:** “Lists support incremental append and delete without replacing the
+> whole list.”
+
+Append a URL to the source list, then delete an entry by index.
+
+### Step 6.3: Set with --report
+
+> **Tell the user:** "`--report` returns JSON with apply status, form state, progress,
+> and remaining issues after each change."
+
+```bash
+markform set /tmp/markform-qa/sp500-research.form.md <url_list_id> --append "https://example.com" --report --format json
+```
+
+### Step 6.4: Confirm Still Complete
+
+```bash
+markform next /tmp/markform-qa/sp500-research.form.md
+```
+
+**Expected:** Form is complete.
+
+## Phase 7: Review the Source
+
+> **Tell the user:** "The filled `.form.md` is both human-readable and
+> machine-parseable. On GitHub, comment tags are invisible: it looks like a clean
+> document. But every value is typed, validated, and programmatically accessible."
+
+```bash
+cat /tmp/markform-qa/sp500-research.form.md
+```
+
+## Phase 8: Present the Report
+
+> **Tell the user:** “Here’s the completed research report.”
+
+Generate the report and render it directly in the chat as formatted markdown:
+
+```bash
+markform report /tmp/markform-qa/sp500-research.form.md
+```
+
+Output the full markdown content to the chat window so the user sees a nicely formatted
+report — headings, tables, lists — with instructions and internal markup stripped away.
+
+Also export the structured data as YAML:
+
+```bash
+markform export /tmp/markform-qa/sp500-research.form.md --format=yaml > /tmp/markform-qa/sp500-research.yml
+```
+
+> **Tell the user:** "The structured data is also saved as YAML at
+> `/tmp/markform-qa/sp500-research.yml` — ready to feed into databases, APIs, or other
+> tools."
+
+Then open the form in a web browser:
+
+```bash
+markform serve /tmp/markform-qa/sp500-research.form.md
+```
+
+> **Tell the user:** "I’ve also opened the report in your web browser.
+> You can browse it in several tabs:
+> - **Form:** the rendered form as it appears on GitHub
+> - **Report:** a filtered view that strips instructions and internal markup
+> - **Edit:** an interactive version where fields can be filled in the browser
+> - **Source:** the raw `.form.md` with all Markform tags visible
+> - **Values:** exported field values as YAML
+> - **Schema:** the JSON Schema describing the form structure
+> 
+> Each tab has a unique URL (hash route) so you can link directly to a specific view."

package/examples/parallel/parallel-research.form.md

@@ -11,12 +11,11 @@ markform:
   role_instructions:
     agent: Research the company and fill in all fields.
 ---
-
 <!-- form id="company_research" title="Company Research (Parallel)" -->
 
 <!-- description ref="company_research" -->
-A company research form that uses  for concurrent deep research
-and  to sequence synthesis after data gathering.
+A company research form that uses `parallel` for concurrent deep research and `order` to
+sequence synthesis after data gathering.
 <!-- /description -->
 
 <!-- group id="overview" order=0 -->
@@ -56,6 +55,3 @@ and  to sequence synthesis after data gathering.
 <!-- /group -->
 
 <!-- /form -->
-
-
-

package/examples/rejection-test/rejection-test.session.yaml

@@ -78,6 +78,19 @@ turns:
         | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
         | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+        ## Incremental Operations (for collections)
+
+        Use these to add/remove items without replacing the entire collection:
+
+        | Type | Example |
+        |------|---------|
+        | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+        | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+        | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+        | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+        | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+        | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
         ## Important: checkboxes vs multi_select
 
         These two types look similar but have DIFFERENT value formats:
@@ -201,6 +214,19 @@ turns:
           | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
           | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+          ## Incremental Operations (for collections)
+
+          Use these to add/remove items without replacing the entire collection:
+
+          | Type | Example |
+          |------|---------|
+          | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+          | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+          | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+          | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+          | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+          | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
           ## Important: checkboxes vs multi_select
 
           These two types look similar but have DIFFERENT value formats:
@@ -381,6 +407,19 @@ turns:
         | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
         | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+        ## Incremental Operations (for collections)
+
+        Use these to add/remove items without replacing the entire collection:
+
+        | Type | Example |
+        |------|---------|
+        | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+        | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+        | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+        | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+        | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+        | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
         ## Important: checkboxes vs multi_select
 
         These two types look similar but have DIFFERENT value formats:
@@ -510,6 +549,19 @@ turns:
           | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
           | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+          ## Incremental Operations (for collections)
+
+          Use these to add/remove items without replacing the entire collection:
+
+          | Type | Example |
+          |------|---------|
+          | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+          | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+          | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+          | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+          | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+          | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
           ## Important: checkboxes vs multi_select
 
           These two types look similar but have DIFFERENT value formats:

package/examples/simple/simple-mock-filled.report.md

@@ -58,8 +58,8 @@ Medium
 
 **Confirmations (Explicit Mode):**
 
-- [x] Data has been backed up
-- [ ] Stakeholders notified
+- Data has been backed up: Yes
+- Stakeholders notified: No
 
 ## URL Fields
 

package/examples/simple/simple-skipped-filled.report.md

@@ -57,8 +57,8 @@ High
 
 **Confirmations (Explicit Mode):**
 
-- [x] Data has been backed up
-- [ ] Stakeholders notified
+- Data has been backed up: Yes
+- Stakeholders notified: No
 
 ## URL Fields
 

package/examples/simple/simple-with-skips.session.yaml

@@ -151,6 +151,19 @@ turns:
         | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
         | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+        ## Incremental Operations (for collections)
+
+        Use these to add/remove items without replacing the entire collection:
+
+        | Type | Example |
+        |------|---------|
+        | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+        | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+        | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+        | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+        | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+        | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
         ## Important: checkboxes vs multi_select
 
         These two types look similar but have DIFFERENT value formats:
@@ -477,6 +490,19 @@ turns:
           | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
           | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+          ## Incremental Operations (for collections)
+
+          Use these to add/remove items without replacing the entire collection:
+
+          | Type | Example |
+          |------|---------|
+          | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+          | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+          | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+          | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+          | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+          | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
           ## Important: checkboxes vs multi_select
 
           These two types look similar but have DIFFERENT value formats:
@@ -979,6 +1005,19 @@ turns:
         | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
         | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+        ## Incremental Operations (for collections)
+
+        Use these to add/remove items without replacing the entire collection:
+
+        | Type | Example |
+        |------|---------|
+        | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+        | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+        | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+        | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+        | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+        | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
         ## Important: checkboxes vs multi_select
 
         These two types look similar but have DIFFERENT value formats:
@@ -1328,6 +1367,19 @@ turns:
           | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
           | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+          ## Incremental Operations (for collections)
+
+          Use these to add/remove items without replacing the entire collection:
+
+          | Type | Example |
+          |------|---------|
+          | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+          | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+          | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+          | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+          | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+          | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
           ## Important: checkboxes vs multi_select
 
           These two types look similar but have DIFFERENT value formats:
@@ -1763,6 +1815,19 @@ turns:
         | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
         | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+        ## Incremental Operations (for collections)
+
+        Use these to add/remove items without replacing the entire collection:
+
+        | Type | Example |
+        |------|---------|
+        | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+        | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+        | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+        | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+        | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+        | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
         ## Important: checkboxes vs multi_select
 
         These two types look similar but have DIFFERENT value formats:
@@ -2085,6 +2150,19 @@ turns:
           | checkboxes | `{ op: "set_checkboxes", fieldId: "tasks", value: { "task1": "done", "task2": "todo" } }` |
           | table | `{ op: "set_table", fieldId: "team", value: [{ "name": "Alice", "role": "Engineer" }] }` |
 
+          ## Incremental Operations (for collections)
+
+          Use these to add/remove items without replacing the entire collection:
+
+          | Type | Example |
+          |------|---------|
+          | append_table | `{ op: "append_table", fieldId: "team", value: [{ "name": "Bob", "role": "PM" }] }` |
+          | append_string_list | `{ op: "append_string_list", fieldId: "tags", value: ["new_tag"] }` |
+          | append_url_list | `{ op: "append_url_list", fieldId: "sources", value: ["https://new.com"] }` |
+          | delete_table | `{ op: "delete_table", fieldId: "team", value: 0 }` (0-based row index) |
+          | delete_string_list | `{ op: "delete_string_list", fieldId: "tags", value: 0 }` (0-based item index) |
+          | delete_url_list | `{ op: "delete_url_list", fieldId: "sources", value: 0 }` (0-based item index) |
+
           ## Important: checkboxes vs multi_select
 
           These two types look similar but have DIFFERENT value formats: