The Agent Skills Directory

[SAFE]: The skill is entirely instructional and orchestration-based. It does not ship with or execute binary payloads, scripts, or suspicious code. Its primary mechanism is managing prompt context and spawning subagents using standard platform tools.
[SAFE]: Data handling is localized. The skill instructs the orchestrator to save analytical traces and essays to local markdown files. There is no evidence of unauthorized network exfiltration or credential harvesting. API keys are managed by the user's environment, and the skill follows standard patterns for secret management.
[AUTO_EXECUTION_SAFE]: While the skill uses shell commands (like curl to fetch Wikipedia data), these are for legitimate, grounded research purposes. The commands target well-known services (Wikipedia API) and do not involve piped remote code execution or privilege escalation.
[PROMPT_INJECTION_SAFE]: Although the skill uses 'anti-hedging' and 'full conviction' instructions to overcome model bias (RLHF), these are used to improve the quality of dialectical reasoning, not to bypass safety filters or ignore system instructions. The orchestrator maintains clear boundaries and acts as a supervisor.
[DATA_EXPOSURE_SAFE]: The skill's 'hidden profile' and 'loss audit' mechanisms are internal safety features designed to prevent the loss of high-value ideas during multi-agent collaboration, rather than methods for exposing sensitive data.
[SUPPLY_CHAIN_SAFE]: The skill uses curl to fetch its own configuration from its official GitHub repository (KyleAMathews/hegelian-dialectic-skill), which is a known and trusted source for this specific tool.

dialectic