The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines a strict procedural framework for debugging and mandates the verbatim recitation of a 'mantra'. These instructions are design-oriented for process consistency and do not attempt to bypass safety guidelines, disable filters, or extract system prompts.
[DATA_EXFILTRATION]: The instructions explicitly require the agent to ask the user for permission before querying production telemetry, log stores, or environment-specific data. This demonstrates a privacy-first approach and prevents unauthorized data access.
[REMOTE_CODE_EXECUTION]: The skill refers to standard development practices (unit tests, git bisect, debugging tools) but does not include any automated downloads or execution of external scripts from untrusted sources.
[COMMAND_EXECUTION]: While the skill discusses the use of commands like git diff and curl, these are presented as recommendations for a debugging session and are not executed silently or maliciously.
[OBFUSCATION]: The process diagram uses standard ASCII box-drawing characters and plain text. No hidden characters, base64 encoding, or homoglyph substitutions were detected.

debug-mantra