What does this skill do, and when should you use it?
Entity Registry is a protocol-layer skill in Aaron Marketing Skills for maintaining the machine-facing identity of one stable entity. It audits and reconciles canonical names, aliases, Schema.org types, Wikidata QIDs, sameAs links, domains, disambiguation evidence, and AI-recognition state. The append-only entity event stream is authoritative, while JSON and Markdown files are rebuildable views. It fits teams managing Knowledge Graph consistency and entity disambiguation, but it does not own brand positioning, voice, or page-level AI-citation optimization.
Reads the entities event stream and projection, Narrative and claims projections, verified source records, and optional rendered views. Through registry-events.py, it queries entities, reviews proposals, and—when the host has the required authority—accepts, rejects, or updates canonical state. It assesses six signal categories: structured data, knowledge bases, NAP+E consistency, first-party content, third-party corroboration, and AI recognition. Each observation records a source, observation date, evidence type, and Pass/Partial/Fail/Unknown status; the workflow can regenerate a Markdown view under memory/entities/ and run verify entities.
- An SEO or GEO team audits an organization’s identity across Schema.org, Wikidata, and AI systems.
- A brand team investigates confusion between its company and a similarly named organization.
- A Knowledge Graph maintainer records a verified Wikidata QID and sameAs set.
- An operations owner reviews pending entity proposals and updates canonical state when authorized.
How do you install this skill?
- Do not accept proposals or write canonical state before confirming the target identity, verified cross-links, lawful basis where applicable, and host write authority.
- Treat knowledge-base, search, and AI-recognition results as uncertain external evidence; the materials do not establish reachability of all relevant services from mainland China.
- Skill version 18.0.0 conflicts with the security policy's supported 17.x line; confirm the actual supported runtime and protocol versions before deployment.
- Shell / CLI
- Local filesystem
Python 3Git
Install it as part of the repository collection. In Claude Code, run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, then /plugin install aaron-marketing@aaron. On an Agent Skills-compatible host, run npx skills add aaron-he-zhu/aaron-marketing-skills; alternatively, run git clone https://github.com/aaron-he-zhu/aaron-marketing-skills.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/aaron-he-zhu/aaron-marketing-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/protocol/entity-registry" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Provide an entity name or aggregate ID, for example: Audit entity recognition for organization acme-analytics. or Review pending entity proposals and reconcile duplicate IDs. Ask for authorization before the first persistent write. Canonical changes require the entity-registry host capability and must be appended through the repository runtime rather than by editing NDJSON or JSON projections directly.
What are this skill's strengths and limitations?
- Clearly separates machine-facing entity facts from narrative, claims, and page-content responsibilities.
- Covers six recognition-signal categories and requires source, date, and evidence-type records.
- Uses an event stream as the canonical source and preserves revision, offset, and dependency information.
- Treats unavailable tools or unobserved engines as Unknown instead of inferring failure.
- Accepting, rejecting, or updating canonical entity state requires an authorized host-capability principal.
- The canonical workflow depends on repository scripts, the event schema, system catalog, and related projections.
- A standalone single-folder installation can prepare a bounded proposal but cannot append, project, or claim canonical truth.
- Natural-person records require lawful-basis checks, data minimization, and erasure controls; the documentation says this is operational guidance, not legal advice.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Entity Registry this page | 60 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Paid Ads Account Auditor | 59 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| TAO DAFT Dataset Validator ✓ NVIDIA · Official | 59 · Recommended | ★ 3.6k | 12d ago | Apache-2.0 |
| AI HOT News Query | 58 · Recommended | ★ 21k | 26d ago | MIT |
| NVIDIA AI-Q Deep Research ✓ NVIDIA · Official | 55 · Use with care | ★ 3.6k | 12d ago | Apache-2.0 |
Its boundary with geo-content-optimizer is explicit: Entity Registry maintains identity and Knowledge Graph facts, while geo-content-optimizer handles page-level AI-citation readiness. Its boundary with narrative-registry is also explicit: Entity Registry owns machine-facing identity, while narrative-registry owns human-facing positioning, naming, voice, and approved descriptions.
How did FollowSkills review this skill?
The evidence shows least-privilege boundaries around a stable non-PII aggregate ID, user confirmation before the first persistent write, host-capability control for canonical mutations, proposal-versus-canonical separation, conflict preservation, and explicit natural-person minimization and erasure safeguards. Five points are deducted because rollback is not explicitly specified, external knowledge-base and AI checks may involve outbound data flows, and the runtime permission and isolation implementation is not fully present in the supplied files.
The evidence shows a defined workflow, decision gates, Unknown handling, revision and offset tracking, verification steps, and controlled stops for missing identity, authority, lawful basis, or conflicts. Two points are deducted because the key scripts, event protocol, runtime, and system catalog implementations are not supplied for reproduction; no skill-specific test coverage or complete failure-message examples are shown.
The evidence shows clear triggers, argument format, entity-type coverage, six diagnostic categories, non-fit boundaries, and handoffs to adjacent registry skills; pasted data and optional tools are supported without requiring paid APIs. Four points are deducted because additional non-fit cases, Chinese-language input/output requirements, and mainland-China reachability or fallback behavior for external knowledge-base and AI services are not specified.
The evidence shows layered documentation covering quick start, contract, boundaries, data sources, procedure, persistence, references, and next skills, plus version 18.0.0, Apache-2.0 licensing, an argument hint, examples, and limitations. Four points are deducted because maintenance ownership, update path, and changelog evidence are unclear in the supplied skill materials; the security policy supports version 17.x while the skill declares 18.0.0, creating a governance inconsistency.
The evidence shows defined audit, reconciliation, proposal, canonical-change, reporting, and handoff outputs for entity identity, schema, sameAs, Wikidata, and AI recognition work; the example report demonstrates a usable output structure. One point is deducted because the example relies on user-supplied AI test results, while actual writes, projection rebuilds, and successful verification are not statically verified, so direct usability still requires review.
The evidence requires source, observation date, evidence type, separate Unknown versus Partial states, event IDs, offsets, revisions, and verification results, with preference for primary sources. One point is deducted because the supplied files contain no third-party execution records, independent reproduction, or dedicated tests covering this skill's key paths, limiting static auditability.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →