The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external web sources, news feeds, and social media posts, which constitutes an ingestion surface for untrusted content.
Ingestion points: Web intelligence and firmographic data retrieved from external providers (Apollo, LinkedIn, Tavily, Exa) via the sandbase_run and sandbase_run_get tools as defined in SKILL.md and references/sandbase-api-map.md.
Boundary markers: The instructions do not specify the use of delimiters or specific markers to separate external research data from the agent's core instructions.
Capability inventory: The skill utilizes the sandbase_run tool to execute capabilities and the sandbase_run_get tool to retrieve results, focusing on data retrieval and synthesis.
Sanitization: There are no explicit instructions provided for sanitizing, escaping, or filtering the content retrieved from external sources before it is synthesized for the user.
[SAFE]: The skill uses vendor-specific tools (sandbase_run, sandbase_inspect, etc.) and established external services to perform its intended research functions. It includes safety guidelines such as a read-only policy and instructions to avoid automatic resubmission of failed tasks.

startup-research