Narrative Registry
Maintain one traceable, versioned source of truth for a brand’s narrative system.
The skill explicitly requires user authorization, expected_revision, host-capability boundaries, append-only events, replayable history, declared data sources, and prohibits direct NDJSON edits. The README and SECURITY material also describe non-persistent credentials, dry-run mutation gates, and private operational memory. Points are deducted because the referenced runtime protocol, scripts, and host deployment boundary are not fully supplied here, so isolation, rollback, and capability enforcement cannot be verified statically.
The contract clearly states read/write paths, atomic complete-canon replacement, stale-proposal rejection, validation, and bounded handoff when runtime capability is unavailable; repository material also indicates regression tests and CI. Points are deducted because execution of registry-events.py, projections, and verify narrative was not reproduced, and concrete abnormal-input diagnostics are delegated mainly to runtime files not provided.
Triggers, inputs, reads, outputs, handoff fields, non-fit cases (TALE scoring and system authoring), and degraded behavior without host capability are explicit. Chinese and English labeling and compatible-host scope are clear, and the core function does not depend on overseas services. Points are deducted because standalone-install boundaries, semantic routing precision, and host reachability in mainland China have limited direct evidence.
Frontmatter, version, Apache-2.0 licensing, Quick Start, shared contract, data sources, numbered instructions, handoff, and next-skill structure are well organized, with references to protocol, security, and registry documentation. Points are deducted because the selected skill does not identify a maintainer, update path, or changelog; repository materials also show a version inconsistency between 18.0.0 and SECURITY’s “17.x current line,” while key dependency documents were not supplied.
The skill specifies querying, recording complete versioned canon replacements, reviewing proposals, emitting exact canon/version pointers, and downstream handoff, so its intended registry task is clearly defined. Points are deducted because no representative executed output or statically verifiable end-to-end result is provided, and core actions depend on runtime and projection scripts outside the supplied evidence.
The selected SKILL.md, README, LICENSE, SECURITY material, CI, and tests provide auditable design evidence, and tests cover shared architecture, audit artifacts, and privacy scanning. Points are deducted because the tests do not establish the narrative-registry key paths, no third-party corroboration or execution result is supplied, and static evidence cannot support a higher score.
- Key dependencies—registry-event-protocol.md, runtime-invocation.md, registry-events.py, system-catalog.json, and verify narrative—were not supplied here and should be checked individually before deployment.
- README identifies 18.0.0, while SECURITY.md calls 17.x the current supported line; confirm the support and security-fix scope for revision ebd4367.
- Verify that the host actually provides the narrative-registry owner capability; otherwise the skill can only leave a proposal and must not claim canonical narrative persistence.
- No scripts, tests, or real event writes were executed for this assessment.
What does this skill do, and when should you use it?
Narrative Registry is the protocol skill for managing a brand’s canonical narrative in Aaron Marketing Skills. It maintains one complete, versioned canon per brand, covering positioning, message hierarchy, voice, naming rules, proof points, and claim pointers. SEO/GEO, social, email, paid, influencer, and launch skills derive their messages from this canon rather than redefining it. It fits teams that need centralized narrative governance with replayable history.
It reads the narrative event stream, projections, accepted positioning and claim evidence, brand voice and naming rules, user-owned examples, and complete proposed canons. It queries the exact current canon version and revision, reviews pending complete proposals, validates internal references and claim IDs, and—when explicitly authorized and the verified runtime, event schema, and system catalog are available—appends events through registry-events.py. Accepted changes regenerate canon.md and versions.md as views. Its handoff records the brand ID, canon version or revision, event ID, superseded version, claim or proof pointers, unresolved contradictions, and one next skill.
- A brand owner needs the current canon version and its proof or claim pointers for a specific brand.
- A strategy team needs to replace an old canon with a complete new version while preserving replayable history.
- Multiple channel teams need one shared source for positioning, message hierarchy, voice, and naming rules.
- A reviewer needs to accept or reject a complete narrative-version proposal that may contain internal contradictions.
What are this skill's strengths and limitations?
- Maintains one complete, versioned narrative canon per brand.
- Preserves old versions in an append-only event stream for replay.
- Covers positioning, message hierarchy, proof pointers, voice, naming rules, and conditional boilerplates.
- Clearly separates narrative curation from TALE scoring, claim-truth adjudication, and narrative-system authoring.
- It does not author the narrative system, score TALE, or adjudicate claim truth.
- Canonical writes depend on a verified runtime, event schema, system catalog, and host capability; standalone one-folder installs cannot append or project.
- The supplied material provides no independent test-suite or per-platform verification evidence for this skill alone.
How do you install this skill?
Install the repository with Claude Code using /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, followed by /plugin install aaron-marketing@aaron. Other Agent Skills-compatible hosts can use npx skills add aaron-he-zhu/aaron-marketing-skills, or you can run git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. This skill is located at protocol/narrative-registry/SKILL.md.
How do you use this skill?
Example prompts include Show current canon version and proof/claim pointers for brand-acme and Record canon v3 as one complete atomic replacement, superseding v2. Explicit authorization is required before writes. Direct canonical mutations must include the current revision and use the host-capability-controlled owner-append flow. If the verified root runtime, schema, catalog, or host capability is unavailable, leave a bounded proposal rather than claiming canonical Narrative truth was written.