Link opportunities

link-opportunities

The authority layer the content engine omits. A perfect map on a zero-authority domain still loses to funded incumbents — this plans the external links (and the citations that travel with them) that make ranking possible. It reads the same brand workspace, respects the tier, and tags every prospect's provenance.

Read first: ../../framework/off-page-authority.md (plays, anchor rules, the honesty line), then ../../framework/eeat-signals.md (why off-site entity consistency matters).

Preconditions

• entity-profile.json (competitors drive the backlink gap) + topical-map.json.
• Prospect discovery needs web_search: true (T1) or dataforseo: true (T2). At T0 the skill produces the plan, plays, and anchor mix but cannot name real prospects — say so; do not invent domains.

Workflow

1. Build the plan skeleton (T0, offline).

   python ../../scripts/link_prospects.py --map brands/<slug>/topical-map.json \
     --entity-profile brands/<slug>/entity-profile.json --brand "<Brand>" \
     --out brands/<slug>/outreach/link-plan.md
   

   Emits, per node (priority-ordered core→outer): the fitting plays, an off-page anchor mix, and a competitor backlink-gap worksheet. Priorities are derived; play notes are asserted.

2. Discover real prospects.

   • T1 (web_search): for top-priority commercial nodes, search the target query + "best/ alternatives/ vs" and record the listicles/reviews that already rank — inclusion targets. For each competitor, search their brand + "review / integration / alternative" to surface sites already covering the category.
   • T2 (DataForSEO backlinks): pull each competitor's referring domains, filter for topical relevance, and rank the gap. Authority numbers here are measured.
   • Record every prospect with: URL, why-relevant, authority (measured or left blank — never guessed), the node it supports, and the play. Score with link_prospects.py:score_prospect (unknown authority scores on relevance + link type, flagged unknown).

3. Write the outreach queue → append to brands/<slug>/outreach/link-plan.md:

   • Ranked prospect table (prospect · node · play · authority[measured|unknown] · angle).
   • Suggested anchor per placement, drawn from the node's anchor examples, honoring the mix (keep exact-match ~10%).
   • No fabricated contacts. If you can't find a real, published contact, mark it "contact: research needed" — do not invent an email.

4. Feed the loop.

   • A high-priority node with no realistic prospects → recommend building a linkable asset for it (free tool / data study), not low-value link spam.
   • Queries where competitors are linked/cited and the brand is absent → shared signal with answer-engine-optimizer.
   • When a link lands, record the URL on the node; it raises that node in seo-performance-tracker.

Definition of done

• link-plan.md written: anchor mix, competitor gap worksheet, priority-ordered plays.
• If T1/T2 on: a ranked prospect queue with provenance on every authority value and no invented domains or emails. If T0: the plan only, with an explicit note that discovery needs web_search/DataForSEO.
• Nothing fabricated — the whole point of doing this inside the suite.

Grounding ladder

• T0: offline plan skeleton, plays, anchor mix, competitor worksheet. Useful alone.
• T1 (web_search): + real listicle/review/resource prospects, measured + dated.
• T2 (DataForSEO): + competitor backlink-gap with measured authority numbers.