Share of Voice Tracker
Measure and trend brand share of voice against a locked competitor panel across platforms and periods.
The skill declares least-privilege data scope, public counts or user exports only, no scraping of closed platforms, and treats fetched content as untrusted input. Memory writes require confirmation, while panel changes preserve the old series and log a break. Deducted 5 points because connector permissions, external data-retention controls, and complete recovery boundaries are not fully specified in this skill; publisher identity is also unverified.
The workflow, formula, locked denominator, break handling, NEEDS_INPUT behavior, and measurement labels are internally coherent, with explicit stop conditions for abnormal cases. Deducted 2 points because key connectors, shared references, and runtime behavior are not supplied here, so static review cannot reproduce calls, rate limits, or failure diagnostics; the score therefore stays below the static ceiling of 10.
Triggers, inputs, expected outputs, non-fit boundaries, and competitor-panel rules are clear, including Chinese-name and misspelling support. Deducted 4 points because platform coverage depends on public data and user exports, geo-relevance is marked low, mainland-China reachability of public connectors is not demonstrated, and some core inputs require manual provision.
The skill has structured frontmatter, Quick Start, Contract, Data Sources, Instructions, and handoff sections; version 18.0.0 and Apache-2.0 are clear, and save/hand-off rules are documented. Deducted 5 points because maintenance ownership, changelog detail, and update responsibility are not clear within the selected skill, while installation and troubleshooting rely mainly on repository-level documentation.
It specifies directly usable SOV tables, locked-panel records, trend break markers, sentiment-weighted variants, and attention-share alternatives, while preventing proxy metrics from being presented as measured. Deducted 9 points because no executed representative output or third-party result evidence is included; connector availability, query coverage, and output quality cannot be verified statically.
The formula, numerator and denominator, platform, period, source labels, and proxy/estimate distinctions provide useful auditability. Deducted 6 points because the evidence is primarily author-authored guidance and repository-internal references, without independently reproducible results, third-party corroboration, or tests covering this skill's key paths.
- This is a source-only review; connectors, queries, and writes were not executed. Static documentation is not evidence of runtime success.
- Public counts, GDELT/Tavily proxies, and Wikipedia attention share are not equivalent to complete social-conversation volume; closed-platform data depends on user exports and as-of dates.
- Before mainland-China deployment, separately verify network reachability, rate-limit behavior, and data compliance for every required public service and connector.
- Changing panel members, query terms, platform sets, or period granularity invalidates the trend and requires a logged break and restarted series.
What does this skill do, and when should you use it?
This skill compares brand and competitor mention counts and calculates share of voice (SOV) by platform and period. It requires a stable competitor panel, query architecture, platform set, and reporting granularity; any coverage change breaks and restarts the trend. It can provide unweighted SOV, a sentiment-weighted variant, and a Wikipedia pageviews-based attention-share alternative. It uses public counts and user exports only and never scrapes closed platforms.
Reads a user-provided or stored competitor panel, brand query variants, exclusions, platforms, and prior panel records. It uses HN, Bluesky, Fediverse, Discourse, GDELT, Tavily, and Wikipedia pageview connectors, while accepting user exports for closed-platform data. It calculates brand mentions divided by brand plus total panel mentions for each platform and period, labeling each cell Measured, User-provided, Estimated, or proxy. Its output includes the locked panel record, a per-platform/per-period SOV table, trend series and break markers, plus requested sentiment-weighted SOV or attention-share results.
- A marketing lead wants monthly SOV comparisons for a brand and three to six fixed competitors across Hacker News, Bluesky, and news echo.
- A social analyst needs to add a competitor while preserving measurement integrity by logging the break, relocking the panel, and restarting the series.
- A brand team wants to compare attention share over the past 12 months using Wikipedia pageviews for its page and competitor pages.
- A marketing team sees an SOV movement that may be a mention spike and hands the result to the social pulse monitoring workflow for investigation.
What are this skill's strengths and limitations?
- Uses explicit numerator, denominator, platform, and period definitions for auditable trend readings.
- Logs panel changes as breaks instead of splicing incompatible time series.
- Distinguishes Measured, User-provided, Estimated, and proxy data and retains unweighted SOV alongside variants.
- Supports public data, user exports, sentiment weighting, and a separate Wikipedia attention-share alternative.
- Closed platforms are never scraped; their data must come from public counts or dated user exports.
- Sentiment-weighted results must be labeled Estimated unless human-coded.
- Changes to panel members, platforms, or query coverage restart the trend.
- The skill owns the SOV number only, not competitor strategy watching, mention sweeps, or backlink analysis.
How do you install this skill?
For Claude Code, add and install the repository plugin with /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, followed by /plugin install aaron-marketing@aaron. On other Agent Skills-compatible hosts, run npx skills add aaron-he-zhu/aaron-marketing-skills, or clone the repository directly. The source does not document a separate installation method for this individual skill.
How do you use this skill?
Provide the brand, locked competitor panel, platforms, and period, for example: Track share of voice for [brand] vs the locked panel [competitor A, B, C] on HN + Bluesky + news echo, monthly, this quarter. To change the panel, ask: We want to add [competitor D] to the SOV panel — log the panel break, re-lock, and restart the series. If no panel is supplied and none is stored, the skill returns NEEDS_INPUT rather than inventing competitors.