The Agent Skills Directory

Command Execution: The skill involves the execution of shell commands, specifically streamlit run, to launch and validate dashboard applications created in Python. This is a standard part of the Streamlit deployment workflow described in the documentation.
Data Access Surface: To build source-backed dashboards, the skill accesses various internal company data lanes, including structured databases, internal documentation, and team communications. This level of access is expected for analytical tools intended to monitor business performance.
Indirect Prompt Injection Surface: Because the skill processes information from external sources like team communications and shared documents, it contains a surface for indirect prompt injection. Maliciously crafted data within those sources could theoretically attempt to influence the agent's dashboard generation or data interpretation logic.
Ingestion points: Data is pulled from ~~company_docs and ~~team_communication as described in SKILL.md (Workflow Step 3).
Boundary markers: No explicit delimiters for untrusted data are defined within the skill's instructions.
Capability inventory: The skill can execute SQL queries, generate HTML/JS visualizations, and run Python-based Streamlit applications.
Sanitization: No specific sanitization or filtering logic is prescribed for the ingested data.

build-dashboard