The Agent Skills Directory

[SAFE]: The skill utilizes a local Python script (resources/macrosensitivity.py) to perform regressions and duration calculations. Technical review of the script confirms it uses only the Python standard library and contains no network operations, file system modifications, or dynamic code execution triggers.
[SAFE]: The instructions in SKILL.md and resource files are strictly limited to financial theory and implementation steps. No patterns of prompt injection, role-play bypasses, or safety guideline overrides were identified.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted financial data for processing.
Ingestion points: Data is read via JSON payloads (payload.json, macro.json, project.json, slopes.json) into the macrosensitivity.py script and the agent's context.
Boundary markers: None explicitly defined in the prompts to separate user data from instructions.
Capability inventory: The agent writes to capital-structure.json and capital-structure.md and executes a local Python script.
Sanitization: The Python script performs input validation, ensuring data points are numeric and the JSON structure is valid, which mitigates simple injection attempts within the calculation engine.

debt-design