The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to collect and store user information, including personal goals, behavioral patterns, and eventually financial data, into local Markdown files (USER.md, NOW.md). This collection is transparent, explicitly stated in the documentation, and follows the primary purpose of creating a persistent user profile. No network exfiltration or unauthorized data access patterns were identified.
[COMMAND_EXECUTION]: The skill performs local file system operations by creating and updating Markdown files in the current working directory. These operations are restricted to standard file writing for context management and do not involve system-level commands, privilege escalation, or persistence via shell configuration files.
[PROMPT_INJECTION]: The skill instructions and its generated operational files (AGENTS.md) define a specific interview protocol and agent personality. These instructions are functional and do not attempt to bypass safety filters or override the underlying agent's core guidelines.
[PROMPT_INJECTION]: There is an attack surface for indirect prompt injection because the agent is instructed to read user-populated files (USER.md, NOW.md) at the start of every session.
Ingestion points: User-provided content in USER.md and NOW.md is read by the agent (defined in assets/AGENTS.md).
Boundary markers: The skill does not implement specific delimiters or 'ignore embedded instructions' markers for the content it writes into the Markdown templates.
Capability inventory: The skill as provided contains no code and defines no high-risk capabilities such as network access, subprocess execution, or arbitrary code evaluation.
Sanitization: No validation or escaping of user input is performed before writing it to the Markdown files.

symbiotic-onboard