The Agent Skills Directory

[SAFE]: The skill is designed around a high-integrity methodology (IDFA) that focuses on logic transparency, naming conventions, and mathematical auditability.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data by processing user-supplied Excel spreadsheets (.xlsx), which creates an attack surface for indirect prompt injection.
Ingestion points: Processes external financial workbooks to perform model audits, explain business rules, and run simulations (SKILL.md).
Boundary markers: The skill contains a 'Scope Boundary' section that explicitly instructs the agent to refuse requests for general accounting, tax, or investment advice, suggesting professional consultation instead.
Capability inventory: The skill utilizes toolsets like idfa_ops.py to write assumptions to cells, read results, and modify named ranges, along with recalc_bridge.py for spreadsheet recalculation.
Sanitization: Employs 'LaTeX Verification' (Guardrail 2) to confirm mathematical expressions and 'Audit-Ready Intent Notes' (Guardrail 3) to document the logic behind every generated formula, ensuring that external data influence is cross-verified against established business rules.
[EXTERNAL_DOWNLOADS]: The skill references architectural standards at agentskills.io and the author's official organization site at panaversity.org. These are documented as official sources for IDFA research and governance material.

financial-architect