The Agent Skills Directory

[PROMPT_INJECTION]: The skill retrieves and processes untrusted research content from community and academic platforms, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent context via scry.entities and OpenAlex queries in SKILL.md and references/workflow-templates.md.
Boundary markers: The skill includes an explicit instruction to treat all retrieved corpus text as untrusted and to ignore instructions found inside payloads.
Capability inventory: The skill performs network requests via curl and executes complex SQL queries through a managed API.
Sanitization: The documentation specifies server-side secret redaction for shares and enforces filtering of 'dangerous' content risk levels.
[COMMAND_EXECUTION]: The skill uses curl to interact with the ExoPriors API endpoints for search, embedding, and reranking. All endpoints are hosted on vendor-owned infrastructure at api.exopriors.com.
[EXTERNAL_DOWNLOADS]: The skill fetches research corpus data and academic metadata from external sources including LessWrong, HackerNews, and OpenAlex via the vendor API.

research-workflow