Domain Authority Auditor Agent Skill

Domain Authority Auditor

Audit one domain's citation-trust signals relative to a declared peer cohort. Produce a complete CITE score only when all applicable evidence and comparison context are present. CITE is advisory; it does not predict whether an answer engine will cite the domain.

When This Must Trigger

• The user asks whether a domain is trustworthy, authoritative, or citation-ready.
• A campaign needs a peer-relative CITE baseline or rerun.
• Link, entity, penalty, or citation evidence may materially change a domain gate.

Quick Start

Audit example.com against these five same-stage U.S. SaaS peers.
Run the CITE content-publisher profile; mark every private-data gap Unknown.

Skill Contract

Use offsite-signal-analyzer for link diagnosis without a gate, content-quality-auditor for one page, and entity-registry for canonical entity repair. Compare multiple domains as separate same-profile runs; never score them as one pooled target.

Reads: one domain, one locked cohort, and dated citation/identity/trust/eminence evidence. Writes: only a permissioned v3 artifact. Done when: context is complete, all 40 states are explicit, the scorer result is preserved, and any saved artifact validates.

Instructions

Runtime Reads

• ../../../references/auditor-runbook.md
• ../../../references/scoring-semantics.md
• ../../../references/cite-domain-rating.md
• ../../../references/framework-catalog.json
• ../../../references/runtime-invocation.md
• references/auditor-runtime.md

Runtime Contract

Read ../../../references/auditor-runbook.md, scoring-semantics.md, cite-domain-rating.md, and the CITE entry in framework-catalog.json. For standalone installs, use bundled immutable references/auditor-runtime.md; never fetch mutable main or guess missing policy. Before deterministic calls, follow runtime-invocation.md, resolve AARON_SKILLS_ROOT="${CLAUDE_PLUGIN_ROOT:-$(git rev-parse --show-toplevel 2>/dev/null || true)}", and require the scorer, validator, and typed catalogs. If unavailable, return score_state: NOT_SCORED / score_confidence: not_scored with no gate verdict or persistent artifact.

Required Setup

Declare:

• one canonical domain and observation date;
• market and entity stage;
• domain type/profile: default, content-publisher, product-service, ecommerce, community-ugc, tool-utility, or authority-institutional;
• a locked peer cohort with inclusion rules and evidence window.

Without peer cohort, market, stage, or domain type, return NEEDS_INPUT/UNDECIDED; the inherited numeric anchors are diagnostic priors, not universal thresholds.

Data Sources

| Need | Preferred evidence |

|---|---|

| Links | Dated backlink export/index with known coverage |

| Identity | Structured data, canonical profiles, knowledge-graph/public entity evidence |

| Trust | Link-history comparison, security checks, authorized console evidence |

| Eminence | Search/analytics visibility and locked peer/query panels |

| Cohort | Named domains plus inclusion/exclusion, market, stage, type, and window |

Evidence Procedure

• Normalize the domain and verify scope (host/subdomain/protocol treatment).
• Freeze backlink, entity, security, search visibility, citation-panel, and reputation evidence with sources/dates.
• Evaluate all 40 IDs against the declared cohort. Tool estimates remain estimated or proxy; user exports remain user-provided unless independently verified.
• Use Unknown for absent backlink indexes, private console data, or comparison universes. Do not mark link-dependent items N/A simply because data is unavailable.
• Apply catalog-authorized N/A only with a reason, for example an AI-citation item when no locked panel is in scope.
• Verify vetoes positively:
• CITE-T03: material link/traffic incoherence plus corroborating manipulation evidence relative to the cohort.
• CITE-T05: verified manipulation network after common-source/ecosystem adjustment.
• CITE-T09: verified active manual action or material unresolved deindexing.

Domain youth, privacy-protected WHOIS, low absolute volume, or missing private-console access cannot by themselves fail a veto.

Create a typed audit-run.schema.json input and execute python3 "$AARON_SKILLS_ROOT/scripts/rubric-score.py" score <run.json> when the verified runtime is available. Never renormalize weights around Unknown items.

Report

Begin with the auditor-runbook's exact typed conversation header. Never replace status, verdict, or score_state with prose; list each explicitly missing qualified item as ID: unknown before findings.

Lead with verdict, profile/cohort/target/date, score or NOT_SCORED coverage/interval, and confidence. Then provide:

• C/I/T/E score and coverage table;
• peer-relative evidence and cohort definition;
• verified risks versus unresolved Unknowns;
• top fixes with owner, evidence needed, and rerun condition;
• provenance appendix, including qualified IDs when detailed traceability is requested.

Use SHIP/FIX/BLOCK/UNDECIDED as the canonical artifact verdict. User-facing prose may say trusted/cautious only as a descriptive translation, never as a separate decision system.

Verdict Rules

• Complete, no veto, healthy score/no failures: DONE + SHIP.
• Complete, remediation needed or one veto: DONE_WITH_CONCERNS + FIX; one veto caps final at 59.
• Complete, 2+ verified vetoes: DONE + BLOCK; no final score.
• Missing applicable/cohort evidence: NEEDS_INPUT + UNDECIDED; no score.

Route link investigation to offsite-signal-analyzer, technical/index evidence to technical-seo-checker, entity repair to entity-registry, and page quality to content-quality-auditor.

§2 CITE Worked Examples

• Complete product-service profile, raw 78, one verified T09 failure: DONE_WITH_CONCERNS/FIX, final 59.
• Complete profile, raw 45, one verified T05 failure: final remains 45; the ceiling never raises a score.
• Verified T03 and T05 failures with complete evidence: DONE/BLOCK, raw retained, no final score.
• Missing backlink universe or private-console evidence: affected items are Unknown and no total is emitted.

§3 CITE Guardrails

• Young domain age and WHOIS privacy are neutral without contradictory evidence.
• A press/launch spike can explain link velocity; test the explanation before labeling manipulation.
• Absolute volume is interpreted within peer cohort, entity stage, market, and domain type.
• Public index absence does not prove a private manual action; T09 requires positive evidence.

§5 CITE Translation

Use plain descriptions by default. On a trace request, show CITE-T03, CITE-T05, and CITE-T09 with source/date/cohort; never expose an unqualified ID.

Persistence

Persist only after explicit authorization. Assemble the complete v3 draft, validate it against the intended memory/audits/domain/YYYY-MM-DD-<topic>.md relative path, persist only through one full-content Write, and revalidate the target. Edit/shell/MCP mutations of the reserved sink are unsupported. Use:

python3 "$AARON_SKILLS_ROOT/scripts/validate-audit-artifact.py" <draft> --relative-path <target>

The audit request itself does not authorize hot-cache, candidate, or registry writes.

Validation Checkpoints

• Target/profile/market/stage/cohort are explicit and locked.
• All 40 IDs have evidence-backed state, Unknown, or valid N/A.
• Absolute anchors were not applied as cohort-independent laws.
• Vetoes are verified failures, not absence-of-evidence shortcuts.
• Scorer output, coverage, confidence, status, and verdict are preserved.
• Persisted artifact is permissioned, path-correct, minimized, and validator-clean.

Reference Materials

• CITE benchmark
• Auditor runbook
• Scoring semantics
• Offsite signal analyzer

Next Best Skill

• Investigate links: offsite-signal-analyzer
• Repair entity state: entity-registry
• Verify technical/index state: technical-seo-checker
• Audit representative content: content-quality-auditor