The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied industry or direction data via the $ARGUMENTS variable. While this is the intended purpose, it lacks explicit boundary markers or sanitization instructions to prevent potential prompt injection from the input influencing the agent's behavior during the multi-step research process. * Ingestion points: User input provided via $ARGUMENTS in SKILL.md. * Boundary markers: None present for the $ARGUMENTS interpolation. * Capability inventory: Executes shell commands (python3 tools/financial_rigor.py), performs web searches, and reads/writes workspace files. * Sanitization: None specified for the input, although the skill mandates cross-checking of the resulting financial data.

era-alpha