What does this skill do, and when should you use it?
Narrative Quality Auditor is the evaluation-stage auditor in Aaron Marketing Skills for brand narrative work. It runs TALE truth, system, and effectiveness profiles separately instead of averaging unlike evidence into one composite score. It checks differentiation, canon alignment, landing-page consistency, and evidence integrity. The skill produces structured audit results and SHIP, FIX, BLOCK, or UNDECIDED verdicts while leaving canon, claims, and surfaces unchanged.
Reads a canon or surface set, a message experiment, and current narrative or claims-truth evidence; declares the target, profile, brand scope, market, audience, canon version, observation date, and evidence window; runs T1–T10 for truth, A1–A10 and L1–L10 for system, or E1–E10 for effectiveness; preserves three separate results in full mode; checks applicable vetoes including TALE-T1, TALE-A1, TALE-L1, and TALE-E1; marks missing evidence as Unknown; and reports status, verdict, score or coverage, confidence, evidence, Unknowns, and fixes. If the required scorer, validator, or typed catalogs are unavailable, it returns NOT_SCORED without a gate verdict or persistent artifact.
- A brand team needs a pre-publish check that its homepage, pricing page, or deck matches the approved canon version.
- A positioning team needs to test whether claimed differentiation is defensible against named alternatives.
- A marketing team needs to review the evidence integrity of a message experiment or resonance claim.
- An organization wants a full narrative review with independent truth, system, and effectiveness results.
- A content team sees message drift across flagship surfaces and needs to locate the consistency failure.
How do you install this skill?
- This is a static-only review; the skill, scorer, validator, and external connections were not executed, so the assessment is not evidence of a successful run.
- The bundled standalone runtime explicitly forbids score calculation and gate verdicts; full behavior depends on the host root runtime and shared references not supplied here.
- Before persistence, verify host permissions, target paths, sensitive-data boundaries, and recovery procedures for audit artifacts.
- Publisher provenance is not verified by the FollowSkills enterprise registry; independently confirm source ownership and maintenance status.
- Shell / CLI
- Local filesystem
Git
The skill is distributed with the Aaron Marketing Skills 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, clone it with git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. The source does not document a host-specific manual installation path for this individual directory.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/aaron-he-zhu/aaron-marketing-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/narrative/evaluate/narrative-quality-auditor" ~/.claude/skills/
rm -rf "$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?
Once installed, send your agent any of these to trigger it:
- audit our brand narrative
- is this message on-canon,
After installation, ask for a narrative audit such as “audit our brand narrative” or “is this message on-canon,” and provide the relevant canon, surfaces, or experiment with versions, market, audience, and dates. The argument hint is <canon/surfaces/experiment> [truth|system|effectiveness|full]. Example: Run TALE system on homepage and pricing page against canon v7 before release. Use full for a complete review and keep the three profile results separate.
What are this skill's strengths and limitations?
- Separates truth, coherence, and effectiveness instead of allowing one evidence type to offset another.
- Treats missing evidence as Unknown and calls out source, date, type, confidence, and required fixes.
- Includes profile-relevant TALE-T1, A1, L1, and E1 veto checks.
- Explicitly prevents autonomous mutation of canon, claims, or surfaces.
- Works with Claude Code and compatible Agent Skills hosts and is Apache-2.0 licensed.
- It is not a launch-readiness audit and does not handle social operations.
- It requires current canon, claims-truth evidence, surfaces, or experiment data; missing inputs can produce UNDECIDED or NOT_SCORED.
- Scoring depends on the scorer, validator, typed catalogs, and reference files named in the source; their full contents are not provided here.
- Effectiveness review requires preregistered comprehension, recall, or behavioral evidence and locked panels; system coherence alone is insufficient.
- The source provides no independent test results or detailed platform-coverage evidence for this individual skill.
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 |
|---|---|---|---|---|
| Narrative Quality Gate this page | 60 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Category Narrative Map | 58 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Audience Belief Mapper | 57 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Narrative Enablement Kit | 55 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Message Test Designer | 61 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
Unlike launch-readiness-auditor, this skill audits narrative truth, system coherence, and effectiveness rather than launch lifecycle readiness. Unlike social-quality-auditor, it does not audit social content or operations. Repairs are handed to positioning-truth-tracer, message-system-architect, narrative-cascade-planner, or message-test-designer.
How did FollowSkills review this skill?
The skill restricts writes to authorized v3 artifacts, requires explicit authorization before persistence, and prohibits mutation of canon, claims, or surfaces. It treats fetched content as untrusted data and specifies fail-closed behavior when runtime prerequisites are unavailable. Deduction: 7 points for relying on host permissions, runtime components, and external references without fully documenting data flows, rollback, or sensitive-data handling in this skill; publisher identity is unknown.
The instructions are internally coherent and define triggers, inputs, outputs, Unknown handling, NOT_SCORED behavior when dependencies are missing, and validation checkpoints. Deduction: 11 points because this is a static review with no execution; the root runtime, scorer, validator, and several referenced files are not supplied, so key-path reproducibility, abnormal-input behavior, and actual failure diagnostics remain unverified.
Trigger semantics, truth/system/effectiveness/full modes, non-fit boundaries, and adjacent-skill routing are clearly declared. The core function does not require an unreachable overseas service, and Chinese labels are included. Deduction: 3 points because required inputs, host capabilities, non-Claude-Code behavior, and Chinese-language output expectations still depend on shared protocols and are not fully specified here.
The skill provides versioning, slug, license, compatibility, Quick Start, Skill Contract, Data Sources, validation checkpoints, persistence rules, and next-skill links in a readable structure. Deduction: 4 points for incomplete maintenance ownership, update path, changelog, and troubleshooting guidance; some referenced resources are absent from the supplied evidence, and publisher provenance is unverified.
The skill separates narrative truth, system coherence, and effectiveness into independently usable results, with explicit evidence, Unknowns, fixes, and no misleading composite score. Deduction: 9 points because no representative execution, audit output, or comparative-benefit evidence is supplied; usefulness depends on unavailable registries, TALE catalog material, scorer, and validator, so direct usability remains unproven.
The supplied immutable runtime snapshot includes a catalog version and source digest, and repository evidence includes architecture and audit-artifact validation tests. Deduction: 6 points because those tests do not demonstrate execution of this skill's key paths, while the benchmark, runbook, schema, and deterministic scorer are not supplied; there is no independent third-party reproduction or corroboration.
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 →