Social Pulse Monitor
Continuously monitor brand mentions, detect spikes, and route each item to the right owner.
The skill explicitly limits itself to read-only listening and routing, prohibits automated posting, replies, DMs, and closed-platform automation, treats fetched content as untrusted, and requires authorized propose events for registry writes with confirmation before promoted writes. Six points are deducted because continuous fetching, Tavily data egress, user exports, and local memory writes still create data-flow and external-effect risks, while rollback and least-privilege boundaries are not fully specified in this skill.
The contract clearly defines inputs, outputs, six-way triage, SLAs, baseline windows, deduplication, proxy labeling, and handling of untrusted content. One point is deducted because connector, shared-protocol, and registry implementations are not included in the skill evidence and key paths were not statically reproduced; static calibration caps this dimension at 10.
Triggers, argument hints, use cases, and boundaries with engagement-inbox-manager, crisis-response-planner, and launch-monitor are comparatively precise, with support for Chinese brand terms and China-specific platforms. Six points are deducted because geo-relevance is marked low, core monitoring partly relies on overseas public services, and mainland-China reachability, fallback sources, and coverage limits lack evidence.
The document has strong structure with frontmatter, Quick Start, Skill Contract, Data Sources, Instructions, save rules, references, version 18.0.0, and Apache-2.0 licensing; repository README, security policy, CI tests, and maintenance signals add useful context. Four points are deducted because the skill lacks an independent sample output, FAQ, changelog, and clearly assigned maintenance/update path, and publisher identity is unverified.
The expected pulse report is concrete and includes a versioned query set, triage table, baseline comparison, spike flag, watchlist hits, and coverage disclosure, giving the core task direct practical value. Nine points are deducted because no representative output or execution evidence is provided; connector reachability, data quality, and routing still require human review. Static calibration caps this dimension at 7.
The skill provides connector, protocol, version, and coverage-disclosure references, while repository tests and CI provide some auditability. Six points are deducted because tests primarily cover shared architecture and audit validation, not this skill's query, fetching, triage, baseline, or write paths; there is no third-party or independent reproduction evidence. Static calibration caps this dimension at 5.
- This is a source-only static review; connectors, queries, triage, baseline calculations, and registry writes were not executed.
- Before mainland-China deployment, verify reachability of GDELT, Tavily, Bluesky, and related services and prepare local or user-export fallbacks.
- Continuous monitoring, third-party search queries, and user exports may involve sensitive brand or business information; confirm authorization, retention, and team SLAs before use.
- Do not treat proxy signals as measured closed-platform data, and do not substitute the default SLAs or 2x-median threshold for account-specific calibration.
What does this skill do, and when should you use it?
Social Pulse Monitor is for always-on brand and community listening. It maintains a versioned query architecture covering brand variants, Chinese names, misspellings, exclusions, and source-specific syntax, then scans public surfaces and triages mentions into six classes. It also maintains a seven-day rolling baseline, flags spikes, and watches B2B accounts for funding, hiring, and launch signals. It fits teams that need public-web monitoring with human follow-up, but it does not handle replies, crisis adjudication, or publishing automation.
Reads user-provided brand terms, product names, executive handles, prior query sets and baselines, plus public-source data or user exports from closed platforms; runs HN, Bluesky, Fediverse, Discourse, GDELT, Tavily and Wikipedia-attention connectors; deduplicates mentions and classifies them as crisis, bug, lead, praise, question or spam with routing targets and SLAs; compares current activity with a seven-day rolling baseline and flags spikes; produces B2B watchlist hits, a coverage-disclosure table and a pulse report. After confirmation, results can be saved to the specified memory path and registry information can be submitted through authorized proposal events.
- A brand team needs continuous monitoring for its brand name, product names, Chinese names and common misspellings.
- A community lead wants to compare this week's mention volume with a seven-day baseline and identify a likely crisis-trigger thread.
- A support team needs to classify new mentions as crises, bugs, leads, praise, questions or spam.
- A B2B sales team wants to watch target accounts for funding, hiring or product-launch signals.
- A team has exports from X, Instagram, TikTok, LinkedIn or Xiaohongshu and wants them reported alongside public-web signals.
What are this skill's strengths and limitations?
- Combines query architecture, public-web sweeps, six-class triage, baseline spike detection and B2B trigger watching.
- Clearly distinguishes Measured, User-provided and proxy data.
- Provides a routing target and SLA for every mention, with defined paths for crises, bugs and leads.
- Designed to work with public or user-provided data without requiring paid tools or API keys.
- X, Instagram, TikTok, LinkedIn, Xiaohongshu and other closed platforms lack compliant keyless read surfaces; their data must be user-exported or proxy-labeled.
- Default SLAs and the spike threshold are Estimated and should be confirmed against account history and team capacity.
- It does not draft or send replies, declare crises or pause queues.
- The supplied source does not provide an independent test suite or platform-coverage validation for this skill.
How do you install this skill?
The skill is located at social/observe/social-pulse-monitor/SKILL.md in the repository. The README documents collection installation: in Claude Code, run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills followed by /plugin install aaron-marketing@aaron; on Agent Skills-compatible hosts, run npx skills add aaron-he-zhu/aaron-marketing-skills; or use git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. The README does not document a standalone installation command for this skill.
How do you use this skill?
Try: "Set up listening for [brand], including misspellings, Chinese names and product names, then build the query set and run today's sweep." Another prompt is: "Run this week's pulse: compare mentions with the seven-day baseline, flag spikes, and give me the triaged routing queue." For B2B monitoring, provide target accounts and request funding, hiring and launch signals. Closed-platform data must be user-exported or clearly labeled as proxy; this skill listens and routes but does not automate posts, replies or DMs.
How does this skill compare with similar options?
Use engagement-inbox-manager for replies, comments and UGC-rights handling; share-of-voice-tracker for competitive trend and share analysis; launch-monitor for T-0 to T+30 launch-window telemetry; and crisis-response-planner for adjudicating a crisis after a spike is flagged.