openclaw/blocked-page-fallback

Blocked Page Fallback

Use this skill when normal web fetch/search is not enough, but the goal may still be reachable through alternate lawful paths.

Do Not Do

• do not bypass login
• do not evade anti-bot or access controls
• do not brute-force endpoints

Fallback Ladder

1. Broaden discovery

• search multiple engines
• use site-specific search
• try alternate titles, aliases, slugs, and locale variants

2. Switch transport

• if plain fetch is thin, use a browser-rendered path
• if browser path is noisy, pivot back to targeted fetch on discovered links

3. Pivot source types

Try allowed alternatives:

• official docs or help centers
• official API or export surfaces
• feeds, sitemaps, changelogs, or release notes
• search-engine cached snippets where available
• public mirrors or archive copies that are openly reachable
• reputable secondary databases

4. Use structural clues

If the exact page is blocked, search by:

• page title fragments
• quoted snippets
• IDs, handles, usernames, product codes, or canonical names
• internal link labels and breadcrumb terms

5. Keep going until confidence is earned

Do not stop after:

• one blocked fetch
• one empty browser render
• one weak search pass

Stop when:

• authoritative or converging sources answer the question
• the remaining blocker is concrete and real
• additional paths are now duplicative

Output Pattern

Return:

1. primary path that failed
2. fallback paths attempted
3. which fallback produced signal
4. best answer now available
5. what would require user-authorized login or a first-party API