Use this skill for practical Airtable management through pydantic-airtable.

Keep SKILL.md focused on workflow. Read references/api-surface.md only when exact method names or signatures matter.

Setup

Install the library if needed:

pip install pydantic-airtable

Provide credentials via environment:

export AIRTABLE_ACCESS_TOKEN="pat..."
export AIRTABLE_BASE_ID="app..."

Most bundled scripts also accept --table and optional --base-id overrides.

Safety notes

• Use a scoped Airtable token with the least privileges needed.
• Prefer a test base before pointing the scripts at production data.
• scripts/model_ops.py --module ./path.py will import and execute that Python module. Only use it with trusted local code.
• --fields, --records, and --updates accept @file.json, which reads local files from disk. Inspect those files before passing them in.
• Installing pydantic-airtable is normal for this skill, but use a virtualenv/container if you want tighter isolation.

Choose the right layer

• Use AirtableManager for base, table, schema, and direct record operations.
• Use AirtableClient for low-level record APIs and batch record operations.
• Use AirtableModel when the user wants typed Pydantic models with CRUD helpers.

For exact method coverage, read references/api-surface.md.

Default workflow

• Confirm credentials are present.
• For one-off operational tasks, prefer the scripts in scripts/.
• For reusable application code, prefer AirtableModel or AirtableManager snippets.
• For schema-sensitive work, validate or sync before writing a lot of data.
• Catch Airtable-specific exceptions around networked operations.

Bundled scripts

scripts/manage_records.py

Use for record CRUD without rewriting the same setup code.

Supported actions:

• list
• get
• create
• update
• delete
• batch-create
• batch-update

Examples:

python scripts/manage_records.py list --table Tasks --max-records 10
python scripts/manage_records.py get --table Tasks --record-id rec123
python scripts/manage_records.py create --table Tasks --fields '{"Name":"Ship it","Status":"Open"}'
python scripts/manage_records.py update --table Tasks --record-id rec123 --fields '{"Status":"Done"}'
python scripts/manage_records.py delete --table Tasks --record-id rec123
python scripts/manage_records.py batch-create --table Tasks --records '[{"Name":"A"},{"Name":"B"}]'

scripts/manage_tables.py

Use for schema and table operations.

Supported actions:

• list-bases
• base-schema
• table-schema
• create-table
• update-table
• delete-table

Examples:

python scripts/manage_tables.py list-bases
python scripts/manage_tables.py base-schema
python scripts/manage_tables.py table-schema --table Tasks
python scripts/manage_tables.py create-table --name Tasks --fields '[{"name":"Name","type":"singleLineText"}]'
python scripts/manage_tables.py update-table --table-id tbl123 --updates '{"name":"Tasks Archive"}'
python scripts/manage_tables.py delete-table --table-id tbl123

scripts/model_ops.py

Use when the task is explicitly model-driven.

Supported actions:

• create-table-from-model
• sync-model
• validate-model

The model file must expose a class by name.

Examples:

python scripts/model_ops.py validate-model --module ./task_model.py --class-name Task
python scripts/model_ops.py sync-model --module ./task_model.py --class-name Task --create-missing-fields

Practical guidance

• Prefer AirtableManager for table creation, schema inspection, and record CRUD across named tables.
• Prefer AirtableClient when the user needs direct record listing or batch writes against a single configured table.
• Prefer AirtableModel when the task should stay type-safe and reusable in app code.
• If the user asks to "manage Airtable" without specifying code output, use the scripts first.
• If a field payload is large or repetitive, write it to a JSON file and pass @file.json to the script arguments that accept JSON input.

JSON input rule for scripts

For --fields, --records, and --updates:

• pass inline JSON like '{"Name":"Task"}', or
• pass @path/to/file.json to load JSON from disk.

Exceptions

Catch and surface:

• ConfigurationError for missing credentials/config
• RecordNotFoundError for missing records
• APIError for Airtable API failures
• ValidationError for model/schema mismatches
• AirtableError as a final fallback