Wellfound

Wellfound platform skill (jessy)

DOM extraction goes through the Claude Chrome extension (claude --chrome). Ask the page in natural terms ("visible job rows", "job detail header", "About the job") rather than relying on fixed selectors.

When this loads

Triggered when the active tab URL contains wellfound.com/jobs, wellfound.com/role, wellfound.com/location, or wellfound.com/remote and jessy is scanning, scoring, or extracting jobs.

URL patterns

Canonical job URL is the detail URL: https://wellfound.com/jobs/<id>-<slug>. Strip query params and fragments.

Page semantics — lists

• Lists are grouped by company. One company card can contain multiple jobs.
• Company block usually has company name, one-line summary, employee band, hiring badges, and funding/stage badges.
• Each job row has: title, employment type, salary/equity when visible, location / remote policy, experience, posted time, save/apply controls.
• Job links often open a full detail page instead of a side rail.
• Lists usually lazy-load more rows. Keep scrolling/loading until no new visible job rows appear, the scan hits an attempted row, or the run cap is reached.

Page semantics — detail

• Header: title, company, salary/equity, location/remote policy, experience, employment type, reposted/posted time.
• Sidebar/details: remote policy, company location, visa sponsorship, preferred timezones, relocation, hiring contact.
• Body: About the job, often with markdown-style headings.
• Company section: summary, employee band, industry/stage/funding tags, perks, founders/team links.

Requirement headings (extract -> req_hard)

Match headings / lead-in phrases, case-insensitive:

• "Requirements"
• "Qualifications"
• "What we're looking for"
• "What you bring"
• "You have"
• "We're looking for"
• "Experience"
• "Skills"

Use nearby bullets/lines until the next heading.

Nice-to-have headings (extract -> req_nice)

• "Nice to have"
• "Bonus"
• "Preferred"
• "Plus"
• "Bonus points"
• "Extra credit"

Split mixed required/preferred lists when visible.

Boundary / skip rules

• Title prefilter: drop cards whose title matches any global skip_title_keywords entry.
• Attempt boundary: if a canonical URL exists in Jessy's attempted state, stop lower/older jobs for that list/feed.
• Ignore Wellfound Save, Apply, and any viewed-like UI state.

Auth + access

Public pages expose enough for many jobs. If the page requires login or blocks detail text, return auth_wall and stop that card. Do not attempt to log in.