Narrative Resonance Monitor
Measures whether a brand narrative is echoed in market language, AI answers, and public signals.
The skill treats pasted data and fetched results as untrusted, labels proxy/Measured/User-provided evidence, keeps the canon read-only, and asks before writing memory. However, it invokes external connectors and writes reports or proposal events without fully specifying skill-level least privilege, rollback, and end-to-end data-flow confirmation, so 7 points are deducted.
Instructions, input/output contracts, NEEDS_INPUT behavior when the canon is missing, method declaration, and termination rules are fairly clear. However, it depends on several connectors and neighboring skills and provides no skill-specific key-path tests, abnormal-output examples, or reproducible execution evidence, costing 2 points within the static ceiling.
Triggers, scope exclusions, argument hints, closed-platform limitations, and expected report structure are explicit. Chinese-market support and mainland-China reachability are not specifically demonstrated, while core signals depend on several overseas services; boundary evidence is therefore incomplete and 5 points are deducted.
The file includes versioning, Apache-2.0 licensing, Quick Start examples, a contract, data sources, save-results rules, references, and handoff guidance. Maintenance ownership and update path are not clear, and SECURITY.md supports 17.x while the skill declares 18.0.0, so 6 points are deducted.
The report schema, evidence labels, echo-rate corpus and method requirements, locked-panel SOV handling, and handoff make the intended output directly useful for narrative evaluation. No representative output or execution verification is supplied, so review and substantial data preparation remain necessary; 1 point is deducted within the static ceiling.
The skill requires declared matching methods, corpora, queries, and as-of dates, and points to repository protocols and security documentation. It lacks skill-specific tests, third-party execution results, and independent cross-source corroboration, costing 1 point within the static ceiling.
- This review is static only; connectors, neighboring skills, and write paths were not executed.
- Reachability, rate limits, regional availability, and response quality of Tavily, Bluesky, GDELT, and Pageviews were not verified.
- The skill declares version 18.0.0 while the security policy lists 17.x as the supported line; confirm permissions, data scope, and recovery procedures before writes or proposal events.
What does this skill do, and when should you use it?
This skill operates in the TALE Evaluate phase to assess whether a durable brand narrative is resonating in the market. It calculates echo rate between market language and the narrative-registry canon lexicon, checks how AI answer engines describe the brand, and measures share of voice on a locked competitor panel. It also collects public signals from Bluesky, GDELT, and Wikipedia pageviews. AI-answer, news, and social results are labeled proxy and are never presented as Measured brand metrics.
Reads the positioning statement, narrative pillars, boilerplate, and language rules from memory/narrative-registry/canon.md; declares the corpus, time window, and matching method before calculating echo rate; uses scripts/connectors/tavily.py --answer to compare answer-engine descriptions with the canon; gathers public resonance or attention signals through gdelt.py, bluesky.py, and pageviews.py; reuses share-of-voice-tracker's locked competitor-panel machinery; incorporates user-provided closed-platform exports; and produces a resonance report labeling echo rate, AI-answer perception, share of voice, and connector signals as Measured, proxy, or User-provided.
- A brand team wants to see whether market mentions are adopting terms from its narrative canon.
- A marketing researcher wants to compare AI-generated brand descriptions with the approved positioning statement.
- A content lead needs to track narrative-related share of voice against a fixed competitor panel.
- A brand owner wants to combine Bluesky, news, and Wikipedia pageview signals into a public resonance read.
- A narrative team needs to hand evaluation findings to narrative-drift-monitor for self-drift and repositioning watch.
What are this skill's strengths and limitations?
- Uses explicit Measured, proxy, and User-provided evidence labels.
- Requires the corpus and matching method to be declared before echo-rate calculation.
- Combines AI-answer perception, locked-panel share of voice, and several public signal sources.
- Has clear scope boundaries: it does not edit the canon or score the final TALE profile.
- Stops without canon.md, because echo rate requires a narrative reference.
- AI-answer, GDELT, and Bluesky results are proxies and cannot stand in for verified audience metrics.
- X, Instagram, TikTok, LinkedIn, and Xiaohongshu have no compliant keyless read; native exports or proxy-labeled data are required.
- The skill references repository scripts and project-memory paths, so standalone porting requires the corresponding file structure and runtime.
- The supplied source provides no standalone test suite or platform-specific test results for this skill.
How do you install this skill?
The skill is located at narrative/evaluate/narrative-resonance-monitor/SKILL.md in the repository. To install the collection in Claude Code, use /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 git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. The source does not document a standalone installation command for this skill.
How do you use this skill?
First provide memory/narrative-registry/canon.md and a locked competitor panel. Example trigger: “Measure narrative resonance for [brand] against our canon lexicon. Competitor panel: [list]. Platforms: [Bluesky / news / all keyless].” For the AI check, use: “Run the AI-answer perception check: how do answer engines describe [brand] vs our positioning statement? Use tavily.py --answer and label it proxy.” Label every number Measured, proxy, or User-provided and include an as-of date; ask before saving results to memory.
How does this skill compare with similar options?
Compared with share-of-voice-tracker, this skill reuses the locked competitor-panel machinery but focuses on narrative resonance and swaps in narrative or message query terms. Compared with performance-monitor, it does not handle own-site GA4/GSC analytics. Compared with narrative-quality-auditor, it supplies evidence for the E dimension but does not score the TALE profile or act as the final gate.