The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's operational instructions include the execution of local Python scripts such as scripts/burn_rate_calculator.py, scripts/unit_economics_analyzer.py, and scripts/fundraising_model.py to generate financial insights.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its requirement to read and process company-context.md before generating output. Malicious instructions placed in this context file could manipulate the agent's behavior during high-stakes financial analysis or script execution.
Ingestion points: Reads company-context.md to identify company stage, ARR, and financial metrics.
Boundary markers: No boundary markers or clear instructions to treat context data as data rather than instructions are present.
Capability inventory: Execution of Python scripts; dynamic task generation for sub-agents using a Task() protocol; invocation of other roles using [INVOKE] syntax.
Sanitization: No validation or sanitization logic is described for data retrieved from external context files.
[NO_CODE]: Several critical functional files mentioned in the documentation, including the Python scripts in the scripts/ directory and supplemental financial playbooks in the references/ directory, are missing from the provided skill structure, precluding a comprehensive security assessment of the actual code being executed.

cfo-advisor