The Agent Skills Directory

[SAFE]: The skill is composed entirely of instructional documentation (Markdown) and structured data templates (JSON). It does not include any executable scripts, binaries, or runtime dependencies.- [DATA_EXFILTRATION]: No sensitive file paths, hardcoded credentials, or suspicious network operations were detected. The skill explicitly provides guidelines on minimizing data collection and redacting sensitive information like credentials and personal data before it enters observability pipelines.- [PROMPT_INJECTION]: The instructions are meta-instructional, guiding the agent on how to perform evaluations. There are no patterns suggesting attempts to bypass safety filters, extract system prompts, or override the agent's core safety constraints.- [OBFUSCATION]: All analyzed files use plain text or standard JSON formatting. No Base64, zero-width characters, homoglyphs, or other obfuscation techniques were found.- [REMOTE_CODE_EXECUTION]: The skill does not perform any external package installations or remote script executions. It relies on the host environment's existing tools and provides framework-neutral methodology.- [DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic shell execution syntax ('!command') was found in the SKILL.md file or other documents.- [PRIVILEGE_ESCALATION]: No commands related to privilege escalation (e.g., sudo, chmod) are present. The skill focuses on verifying authorization and permissions within the agent's task trajectories.

agent-evals-and-observability