The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves fetching and processing external, untrusted content such as company filings (10-K, 10-Q) and investor presentations. This creates an indirect prompt injection surface where an attacker could embed malicious instructions within public documents to override agent behavior or bias the financial verdict.
Ingestion points: SKILL.md Workflow Step 1 ("Pin the company and pull the primary documents") and Workflow Step 2 ("get the latest annual report").
Boundary markers: Absent; there are no instructions to the agent to treat fetched content as data only or to ignore embedded natural language instructions within the filings.
Capability inventory: The agent is authorized to execute local shell commands and Python scripts (scripts/quality_check.py and scripts/reverse_dcf.py).
Sanitization: Absent; the skill relies on the agent's internal capability to interpret filing text safely.
[COMMAND_EXECUTION]: The skill instructs the agent to execute bundled Python scripts to perform financial modeling and trend analysis.
Evidence: SKILL.md contains explicit instructions to run python3 scripts/quality_check.py for earnings quality checks and python3 scripts/reverse_dcf.py for valuation calculations.
[EXTERNAL_DOWNLOADS]: The skill requires network access to fetch financial data and primary documents from various external sources.
Evidence: SKILL.md requires "Live access to primary sources" and instructions to "search or fetch filings" from company investor relations sites and financial data aggregators.

stock-deep-dive