The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill aggregates data from several external financial platforms and supports manual uploads of PDF/CSV statements. This ingestion of untrusted external content introduces a surface for indirect prompt injection, where instructions hidden in the data could attempt to manipulate the agent.
Ingestion points: Step 1 and Step 2 involve pulling data from iCapital, Addepar, Orion, Envestnet, and Google Drive, alongside user-uploaded files.
Boundary markers: The instructions do not mandate the use of specific delimiters or protective wrappers around external data to isolate it from the agent's core instructions.
Capability inventory: The skill possesses the capability to invoke subagents (source-extract), perform tool searches, and generate files (Markdown, .docx, .pdf).
Sanitization: No explicit sanitization or filtering logic is prescribed for the data retrieved from external sources before it is processed or presented.
[Sensitive Data Handling]: The skill accesses detailed financial records, including NAVs, commitments, and account identifiers. It implements a 'confirm identity' rule to ensure data is retrieved for the correct entity, which serves as a functional control against accidental data exposure.
[Automated Task Execution]: The skill utilizes parallel subagent calls to gather data efficiently. While this optimizes performance, it relies on the safety profiles of the underlying subagents and the platform's tool-registry security.

alts-brief