The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is instructed to 'Always read company-context.md before responding', which creates an ingestion point for untrusted external data. If this file contains malicious instructions, they could influence the agent's reasoning or the parameters it uses when executing its analytical scripts.
Ingestion points: company-context.md (referenced in SKILL.md).
Boundary markers: None identified; the skill lacks instructions to delimit or ignore instructions embedded within the context data.
Capability inventory: The skill includes and executes two local Python scripts (scripts/strategy_analyzer.py and scripts/financial_scenario_analyzer.py) which process data and provide strategic recommendations.
Sanitization: No explicit sanitization, validation, or filtering of the external content is prescribed before the data is processed by the agent or the scripts.

ceo-advisor