The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from external, untrusted sources which creates a potential surface for indirect prompt injection attacks.
Ingestion points: As described in Phase 2 of SKILL.md, the agent is instructed to fetch and analyze content from "SEC filings" and "Web search for current quotes, multiples, and historical data".
Boundary markers: There are no explicit delimiters or instructions to the agent to disregard instructions potentially embedded within the ingested external data.
Capability inventory: The skill allows writing to local HTML files (Phase 5) and executing local scripts (Phase 4).
Sanitization: No specific sanitization or validation logic is provided to handle the external data before it is rendered into the final HTML output.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md (Phase 4) and documentation in references/financial-components.md direct the agent to execute a local script named chart_generator.py to generate visualization assets (waterfall and football-field charts).

ib-deck