The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes platform capabilities to invoke subagents for literature searches and delegates data discovery tasks to an external 'find-data' skill.
[DATA_EXFILTRATION]: The skill writes output in the form of markdown research briefs to the user's local directory (e.g., 'research_briefs/'). This is a standard functional behavior for generating and saving the results of a brainstorming session.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes various forms of untrusted external content. Ingestion points: In Phase 2, the agent reads user-provided CVs, papers, or project directories; in Phase 4, it processes results from automated academic searches; in Phase 8, it ingests data from external skill outputs. Boundary markers: There are no specified delimiters or 'ignore' instructions mentioned to distinguish external data from the skill's internal logic. Capability inventory: The agent has permissions to write to the file system and call other tools or skills based on the processed inputs. Sanitization: The instructions do not outline any sanitization or validation protocols for the external documents or search results before they are integrated into the session context.

research-brainstorm