The Agent Skills Directory

[SAFE]: The skill is built with a security-first approach, incorporating a dedicated 'Context Engine' and an 'Anonymization Protocol'. These systems are specifically designed to scrub sensitive business information (such as ARR, runway, and customer names) from being sent to external APIs or search engines, ensuring data privacy during strategic analysis.- [COMMAND_EXECUTION]: The skill includes a suite of 25 Python-based analytical tools (e.g., burn_rate_calculator.py, risk_quantifier.py). These tools are static, run locally, and exclusively use Python standard libraries for mathematical modeling and data parsing, with no evidence of dangerous system calls or network exfiltration.- [PROMPT_INJECTION]: The skill implements a 'Board Meeting Protocol' and an 'Agent Protocol' that define strict communication rules for multi-role deliberations. These protocols include 'Phase 2 Isolation' to prevent cross-contamination of thoughts and an 'Executive Mentor' role to act as a critic, ensuring robust and safe agent behavior.- [DATA_EXFILTRATION]: No data exfiltration vectors were identified. The skill manages company context and decision logs in a controlled local environment (~/.claude/), and the instructions provide explicit guardrails against the external transmission of non-anonymized proprietary data.

c-level-advisor