The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions that attempt to redirect the agent's authority to an external file located two directories above the skill's root. By designating ../../SKILL.md as the 'authoritative workflow', it seeks to override its own instructions and potentially escape the constraints of its specific directory.
Evidence: "Read ../../SKILL.md completely and follow it as the authoritative workflow."
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data formats, including financial datasets, spreadsheets, and PDFs, which are known vectors for embedded malicious instructions.
Ingestion points: Revenue forecasts, SaaS/RevOps datasets, spreadsheets, and PDFs as specified in the skill description in SKILL.md.
Boundary markers: The skill instructions do not provide delimiters or instructions to ignore prompts embedded within the processed data.
Capability inventory: The skill description implies the use of analytical, visualization, and data processing tools.
Sanitization: No evidence of input validation, filtering, or sanitization protocols for the processed files is provided.

data-analysis