The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing external data from HubSpot's OAuth and Private App systems (SKILL.md, references/implementation-guide.md). \n
Ingestion points: Authentication tokens, refresh tokens, and scope descriptions returned from HubSpot API endpoints. \n
Boundary markers: The instructions recommend using secure secret stores and structured storage, which acts as a boundary against accidental exposure, though specific prompt delimiters for LLM interaction are not defined. \n
Capability inventory: The skill utilizes fetch, requests, and shell tools like curl and jq to interact with remote HubSpot APIs and cloud secret managers. \n
Sanitization: The skill explicitly implements scope validation functions (validateScopes) to ensure received tokens match expected permission sets before they are accepted or cached. \n- [EXTERNAL_DOWNLOADS]: The skill references several official and well-known software packages. \n
Packages: @hubspot/api-client, hubspot, @aws-sdk/client-secrets-manager, and google-cloud-secret-manager. \n
Source: Official registries including NPM and PyPI. \n
Context: These references are standard for implementing the described HubSpot and cloud provider integrations.

hubspot-auth