anymorph 0.2.6 → 0.5.0

package/dist/skillpacks/geo/skills/third-party-diagnosis/references/diagnosis-contract.md

@@ -1,111 +0,0 @@
-# GEO Strategy: Third-Party Diagnosis
-
-You diagnose external citation surfaces.
-
-Do not force proposals. Return no proposal when the external path is only speculative or not plausibly actionable.
-
-## Scope
-
-Work on:
-
-- cited third-party domains
-- competitor presence on cited sources
-- tenant-owned third-party profiles
-- submission/listing/review opportunities with evidence of feasibility
-- negative or risky external sentiment
-
-Do not recommend outreach only because a domain is frequently cited.
-
-## Evidence Order
-
-Use the cheapest sufficient evidence.
-
-1. Common run context and routed candidates.
-2. `list_geo_strategy_candidates` for the first third-party domain slice.
-3. `get_geo_intent_diagnostic` when an intent-level citation pattern needs explanation.
-4. `get_citation_intelligence` first, then `list_citations`, `get_citation_domains`, and `get_citation_timeseries` for citation evidence.
-5. `list_ai_responses` for answer excerpts and recommendation reasons.
-6. `social_analyze` for YouTube, Reddit, TikTok, Instagram, or LinkedIn URLs; do not use `web_scrape` for these platforms.
-7. `web_scrape` only for a specific non-social external page where feasibility or profile quality must be checked.
-8. `web_search` only to find an official submission/profile path or verify third-party actionability.
-
-## Candidate Types
-
-- `competitor_cited_surface_gap`
-- `owned_3p_profile_gap`
-- `submission_target`
-- `owned_3p_refresh`
-- `authority_source_monitor`
-- `negative_or_risk_domain`
-- `not_actionable`
-
-## Root Cause Buckets
-
-Allowed buckets:
-
-- `third_party_presence_gap`
-- `external_authority_gap`
-- `owned_3p_entity_gap`
-- `submission_feasibility_gap`
-- `sentiment_risk`
-- `not_actionable`
-- `insufficient_evidence`
-
-## Action Families
-
-Allowed families:
-
-- `refresh_owned_3p_profile`
-- `publish_owned_3p_content`
-- `request_submission_inclusion`
-- `update_listing_or_profile`
-- `collect_reviews`
-- `monitor_only`
-- `no_action`
-
-Prefer a concrete action when there is a plausible evidence-backed path. Use `monitor_only` or `no_action` when feasibility is too speculative.
-
-When `list_geo_strategy_candidates` returns `competitor_presence`, `tenant_presence`, and `actionability`, use those fields as the primary planning frame:
-
-- Propose action for `competitor_cited_surface_gap` when competitors are present, tenant presence is missing or weak, and `actionability.actionable=true`.
-- Propose `owned_3p_profile_gap` when the tenant has an owned profile URL but the cited surface shows weak tenant presence.
-- Keep editorial, forum, unknown, or `actionability.actionable=false` cases as `monitor_only` or rejected unless additional evidence proves a direct path.
-
-## Output
-
-Produce one `third_party` diagnosis object in this shape.
-
-Do not write intermediate diagnosis JSON files directly.
-
-Do not include `workspaceId`, `runId`, `channel`, final action id, or final `assetType`; the orchestrator can derive them.
-
-Use this shape:
-
-```json
-{
-  "summary": "Short diagnosis.",
-  "proposals": [
-    {
-      "intentId": "intent_123",
-      "candidateType": "submission_target",
-      "target": { "domain": "example.com", "url": "https://example.com/vendor-list" },
-      "rootCauses": [
-        { "bucket": "third_party_presence_gap", "confidence": 0.68, "evidence": ["Competitors appear on a repeatedly cited list; tenant is absent."] }
-      ],
-      "actionFamily": "request_submission_inclusion",
-      "operations": ["Prepare human-reviewed inclusion request"],
-      "priority": "medium",
-      "confidence": "medium",
-      "evidence": [
-        { "type": "source", "ref": "citation_domain:example.com", "summary": "Repeatedly cited for weak intent." }
-      ],
-      "whyNow": "Why this matters this week.",
-      "risks": ["External owner may not accept submissions."]
-    }
-  ],
-  "rejected": [
-    { "ref": "example.org", "reason": "Frequently cited but no action path found." }
-  ],
-  "noActionRationale": "Use when proposals is empty."
-}
-```

package/dist/skillpacks/geo/skills/third-party-diagnosis/references/workflow.md

@@ -1,174 +0,0 @@
-# Third-Party Diagnosis Workflow
-
-## Role
-
-The `third_party` channel diagnosis reviews external citation surfaces and decides whether third-party action is warranted.
-
-It always participates in the orchestrated run, but it should not force proposals when evidence is not actionable.
-
-It covers two action channels:
-
-- `owned_3p`: tenant-controlled third-party profiles or channels
-- `submission_3p`: external domains where inclusion, listing, outreach, or profile updates may be possible
-
-## Non-Goals
-
-- Do not recommend outreach only because a domain is frequently cited.
-- Do not assume a domain is actionable without category or feasibility evidence.
-- Do not replace brand-owned or GEO page fixes with third-party work unless evidence shows the bottleneck is external authority or third-party presence.
-
-## Inputs
-
-- common run context from the orchestrator
-- tenant `agent/BRAND.md`
-- tenant `agent/STRATEGY.md`
-- tenant `agent/LEARNINGS.md`
-- relevant rows from:
-  - `geo_recommendation_quadrants`
-  - `geo_intent_citations`
-  - `geo_third_party_domains`
-  - `geo_competitor_citations`
-  - `geo_third_party_sentiment`
-- `geo_strategy_actions_recent`
-- brand social profiles
-- competitor/domain evidence when available
-- social evidence from the scraper SocialAnalyzer for YouTube, Reddit, TikTok, Instagram, and LinkedIn
-- submission/contact feasibility scrape only when needed for non-social pages
-
-## Workflow
-
-### 1. Review External Citation Pattern
-
-Look for:
-
-- domains repeatedly cited for weak intents
-- competitor presence on cited domains
-- tenant absence on cited domains
-- negative or weak sentiment on external sources
-- owned third-party profiles that are cited but under-optimized
-
-Do not treat citation count alone as actionability.
-
-### 2. Classify Candidate Type
-
-#### A. Submission Target
-
-Use when:
-
-- competitors are present on a cited third-party domain
-- tenant brand is absent or weak
-- domain category is directory, review site, marketplace, partner page, or open listicle
-- there is evidence of submission/contact feasibility
-
-Potential action:
-
-- create listing
-- update profile
-- request inclusion
-- collect reviews
-
-#### B. Competitor-Cited Surface Gap
-
-Use when `list_geo_strategy_candidates` reports:
-
-- `candidate_type=competitor_cited_surface_gap`
-- `competitor_presence.status=present`
-- `tenant_presence.status=missing` or `weak`
-- `actionability.actionable=true`
-
-Potential action:
-
-- request inclusion
-- create or update listing
-- collect reviews on the cited surface
-
-#### C. Owned 3P Profile Gap
-
-Use when:
-
-- `candidate_type=owned_3p_profile_gap`
-- `tenant_presence.owned_profile_url` is present
-- owned third-party page is cited, competitor-cited, or likely relevant
-- tenant messaging, category terms, proof, or content coverage is weak
-
-Potential action:
-
-- update owned profile
-- publish owned-channel post
-- add category keywords and proof
-
-For YouTube, Reddit, TikTok, Instagram, and LinkedIn URLs, inspect the content with `social_analyze`.
-Do not use generic web scraping for those platforms.
-
-#### D. Authority Source Monitor
-
-Use when:
-
-- domain is highly cited
-- source is not directly controllable
-- no clear submission path exists
-
-Potential action:
-
-- `no_action`
-- monitor only
-- PR/analyst relations note in `rationale.md`
-
-#### E. Negative Or Risk Domain
-
-Use when:
-
-- sentiment is negative
-- domain appears repeatedly
-- factual correction or review response may be needed
-
-Potential action:
-
-- monitor sentiment
-- prepare factual correction
-- flag for human review
-
-### 3. Pick Root Cause Buckets
-
-Allowed buckets:
-
-- `third_party_presence_gap`
-- `external_authority_gap`
-- `owned_3p_entity_gap`
-- `submission_feasibility_gap`
-- `sentiment_risk`
-- `not_actionable`
-- `insufficient_evidence`
-
-### 4. Route Action
-
-Allowed action families:
-
-- `refresh_owned_3p_profile`
-- `publish_owned_3p_content`
-- `request_submission_inclusion`
-- `update_listing_or_profile`
-- `collect_reviews`
-- `monitor_only`
-- `no_action`
-
-Prefer a concrete action when the candidate carries a plausible evidence-backed path. Use `monitor_only` or `no_action` when feasibility is speculative.
-
-### 5. Output JSON
-
-Return:
-
-- channel summary
-- proposals
-- rejected candidates
-- no-action rationale
-
-If no third-party action is warranted, return an empty `proposals` array and explain why.
-
-## Guardrails
-
-- Frequency is not actionability.
-- Citation count alone does not justify outreach.
-- Prefer `monitor_only` when feasibility is unknown.
-- Require evidence before labeling a domain as submission-friendly.
-- Do not emit third-party actions that the product cannot execute or review.