The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to initialize the environment using uv run python -m src.cli.instance_init and to perform data readouts via uv run python -m src.integrations.refresh_all. It also uses cp to move user-provided files.
[DATA_EXFILTRATION]: The skill accesses sensitive financial data, including broker-exported CSV files and a user profile (user-profile.yaml) containing risk tolerance and account details. Although the skill instructions explicitly forbid echoing this data or moving it out of the local instance, the capability to read and process these sensitive files is a core part of the workflow.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by inspecting the headers of user-supplied broker CSV files to identify export types.
Ingestion points: Header inspection of the <broker-export.csv> file (SKILL.md).
Boundary markers: None specified; there are no specific delimiters or warnings to ignore instructions that might be embedded in the CSV file metadata or headers.
Capability inventory: The skill can execute shell commands (uv, cp, cd) and modify local configuration files (user-profile.yaml).
Sanitization: No sanitization or validation logic is defined for the content of the CSV headers before the agent inspects them.

instance-onboarding