The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for ingesting and processing data from external sources such as visitor identifiers from cookies and metadata from third-party webhooks. (1) Ingestion points: Visitor IDs from client-side cookies and metadata returned via webhooks from third-party platforms in references/first-party-tracking.md. (2) Boundary markers: The skill explicitly instructs the implementation of validation logic such as isAnonymousDistinctId to ensure data is anonymous. (3) Capability inventory: Examples include network requests using fetch and identity management via identify. (4) Sanitization: The provided code snippets demonstrate sanitization through trimming, case normalization, length constraints, and explicit PII detection (e.g., checking for "@" symbols).
[DATA_EXFILTRATION]: The skill describes methodologies for transmitting identifiers to marketing and analytics platforms like PostHog and SavvyCal. These operations target well-known services and include specific instructions to implement an 'anonymity guard' to prevent the exposure of PII or identified user data in URL parameters or webhooks.

marketing-attribution