Elysia Tools
Navigation
Workflow Playbook
Turn mixed text sources into reviewable links, attributes, dates, log fields, and sensitive-match inventories with a deliberate extraction path.
Hubs
This workflow is for people who need a checkable set of signals from raw text, HTML, Markdown, or logs. Before opening a tool, preserve the source and write the expected fields, output shape, duplicate policy, and the context a reviewer will need. A small positive and negative fixture is more useful than a vague instruction to extract everything.
Use text-extractor for a first pass when the fields are not yet settled. If the task is already a known URL collection, bulk-url-extractor gives the task a narrower boundary. Do not flatten HTML before deciding whether attributes or image sources are part of the result: use html-attribute-extractor or image-source-extractor first, then create clean reading text with new-html-tag-stripper if that derivative is needed. Markdown links deserve markdown-link-extractor because link syntax carries information that plain text can obscure.
For logs, ip-address-extractor and date-extractor provide candidate fields, but the match is only reviewable when its original line, nearby event text, and timezone assumption stay attached. Compare the extracted count and representative values with the fixture. A matching IP is still just an extracted value, not an identity conclusion, and a date string is not automatically a business event.
If the input may contain contact or network identifiers, decide authorization and minimum scope before using bulk-email-extractor, phone-number-extractor, or ip-address-extractor. Restrict raw outputs, record the fields and rows released, and mask or pseudonymize the handoff when required. Accept the result only when the expected fields are present, unwanted matches are explained, source context is preserved, and the delivered artifact follows the handling policy. The extraction tools do not replace a redaction or access-control process.
Workflow playbook
Save an untouched copy and write down the fields you expect. Use `text-extractor` for a broad first pass over mixed text, or use `bulk-url-extractor` immediately when the task is specifically to collect HTTP and HTTPS links. Keep the baseline so later targeted results can be compared with it.
For HTML, run `html-attribute-extractor` or `image-source-extractor` before `new-html-tag-stripper` when attributes, links, or image sources matter; use the stripper only for a clean reading-text derivative. For Markdown, use `markdown-link-extractor` to retain link syntax and then review the extracted list separately from the readable text.
Use `ip-address-extractor` for IPv4 and IPv6 candidates and `date-extractor` for dates in mixed records. Retain the original line, surrounding event context, and timezone assumption so a match can be checked without claiming that the IP identifies a person or that a date has a particular business meaning.
Use `text-extractor` for a quick mixed-pattern pass when the fields are still being discovered. Choose `bulk-url-extractor` or another targeted extractor when the field, deduplication rule, and output purpose are already known.
For HTML, extract attributes or image sources before removing tags if markup carries meaning. For Markdown, use `markdown-link-extractor` so inline, reference, and bare links are not confused with ordinary text. Treat logs as line-oriented records rather than as a single paragraph.
Extract email, phone, or IP fields only with authorization and only at the minimum scope needed. Keep raw values access-controlled, produce masked or pseudonymized handoff data when required, and do not treat an extractor as an anonymization or compliance control.
When the authorized task includes contact or network identifiers, use `bulk-email-extractor`, `phone-number-extractor`, and, where needed, `ip-address-extractor` to build the smallest required inventory. Compare counts with a fixture, restrict raw values, and mask or pseudonymize the handoff; these tools identify candidates but do not perform automatic anonymization.