The Agent Skills Directory

[PROMPT_INJECTION]: No patterns of instruction override, jailbreak attempts, or system prompt extraction were found. The skill maintains clear scope boundaries and redirects out-of-scope requests to appropriate specialized agents.
[DATA_EXFILTRATION]: The skill does not perform network requests or access sensitive local files. It includes explicit rules to exclude or redact PII and secrets from telemetry designs.
[REMOTE_CODE_EXECUTION]: There are no remote script downloads, package manager invocations, or dynamic code execution patterns (like eval or exec) present in the skill.
[COMMAND_EXECUTION]: The skill provides high-level architectural guidance and does not involve the execution of shell commands, subprocesses, or dynamic context injection.
[OBFUSCATION]: All files were scanned for Base64, zero-width characters, homoglyphs, and other encoding techniques; no hidden or deceptive content was identified.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private secrets were found in the instructions, references, or metadata.

observability-advisor