The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is intended to process external business plan files (PDF, PPTX, and Markdown). While this creates an ingestion point for untrusted data, the skill includes explicit 'Output Contract' and 'Protect facts' instructions that mandate the separation of business facts from assumptions and forbid the fabrication of data, which serves as a safety boundary for the agent's reasoning.
[COMMAND_EXECUTION]: The skill executes a local script scripts/audit_bp.py to perform structural linting. This script uses standard Python libraries to parse text and calculate scores based on structural 'beats'. It operates within the provided project workspace and does not perform network operations or access sensitive system paths.

lov-bp-polish