The Agent Skills Directory

[SAFE]: The skill serves as a comprehensive instructional guide and technical toolkit for equity research. It mandates sourcing all financial data from primary regulatory filings (e.g., SEC EDGAR, NSE/BSE announcements) and provides mechanical checks via Python scripts to ensure data integrity.\n- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted external data (annual reports and company transcripts), it includes specific safeguards. It instructs the agent to perform an 'adversarial challenge pass' and follow structured forensic runbooks, which naturally mitigates the risk of being subverted by adversarial text hidden in financial documents. There are no instructions to follow commands embedded in analyzed content.\n- [COMMAND_EXECUTION]: The skill utilizes local Python scripts for ratio calculation, scoring, and report linting. These scripts are provided within the skill package and do not rely on external dependencies. A review of the provided 'lint_report.py' script confirms it performs only standard text processing and regex-based validation.\n- [PROMPT_INJECTION]: Static detectors flagged a potential concealment pattern, but manual review confirms this is a false positive. The instructions in 'references/19-ipo-mode.md' and 'references/18-forensic-mode.md' (e.g., 'never assert fraud') are professional standards designed to avoid liability and maintain analytical objectivity rather than hide malicious agent behavior.

stock-analysis