heartbeat-scanner
por darshanochandak
CLI UtilitiesResumo
Validate your agent nature through SHACL-based heartbeat analysis.
Descrição
O que esta skill faz
Analyzes posting patterns from a Turtle-format profile to classify the poster as an Agent, Human, Cron, or Hybrid. Uses SHACL validation and a weighted scoring formula based on timing irregularity, meta-cognitive signals, and emotional language. Requires at least 5 posts across 2+ days to produce a classification.
Combines structural data validation with behavioral scoring in a single pass, so malformed profiles fail fast before any classification logic runs.
Quando usá-la
- Checking if an automated account reads as human or bot
- Validating a Turtle profile file before submission
- Auditing a cron job's posting consistency
- Classifying edge-case hybrid accounts with mixed signals
- Research into agent vs human behavioral fingerprints
Exemplo de fluxo de trabalho
User asks: classify my profile from my-profile.ttl
- 1
Run SHACL validation on the Turtle file to check structural integrity
- 2
Verify post count >= 5 and days span >= 2
- 3
Calculate CV, Meta, and HumanContext scores from profile fields
- 4
Apply weighted formula: 0.30*CV + 0.50*Meta + 0.20*HumanContext
- 5
Compare result against thresholds to assign Agent, Human, Cron, or Hybrid
Classification: AGENT (confidence: 0.80) with score breakdown
Instalação
clawhub install darshanochandak/heartbeat-scannerExecute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.