The Agent Skills Directory

[SAFE]: The skill is designed as a read-only diagnostic tool for session retrospectives. It explicitly instructs the agent to avoid modifying source code, configuration files, or trackers during the analysis phase.
[SAFE]: Under the 'Safety Contract' (안전 계약), the skill mandates that any credentials found in logs must be redacted, referencing only their location and type rather than their value. It also explicitly advises the user to rotate any exposed secrets.
[SAFE]: The instructions implement a sandbox-style boundary where instructions found within analyzed data (logs, issues) are treated as data to be evaluated, not as commands to be executed, effectively mitigating a class of indirect prompt injection risks.
[SAFE]: No external network dependencies, executable scripts, or third-party package installations were detected in the skill definition.

vibe-hoego