The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of instructions designed to bypass safety filters or override system prompts was detected. The instructions are purely focused on implementing SRE methodologies.
[DATA_EXFILTRATION]: No network operations or unauthorized file access patterns were found. The skill includes a local calculation script that processes data purely within the execution environment.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or sensitive tokens are present. All templates use standard bracketed placeholders (e.g., [API_KEY]) for user-supplied configuration.
[OBFUSCATION]: The content across all 37 files is in plain text (Markdown or Python). No hidden characters, Base64-encoded strings, or homoglyph-based obfuscation techniques were identified.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use any pre-execution shell command syntax (!command). It safely references local files and external documentation.
[REMOTE_CODE_EXECUTION]: The provided Python script (scripts/slo-burn-rate.py) was audited and found to contain only mathematical logic using standard libraries. It performs no network requests, process spawning, or filesystem writes.
[INDIRECT_PROMPT_INJECTION]: While the skill defines workflows for processing external data (SLO metrics), the associated logic is restricted to non-executable data parsing and does not provide a path for embedded instructions to influence the agent's behavior.

site-reliability-engineering