The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were found during the analysis. The skill's behavior is consistent with its stated purpose of auditing observability practices.
[COMMAND_EXECUTION]: The skill uses the Bash tool and local scripts (cli.mjs) to manage audit state and execute analysis tools like linters or type checkers. This is an expected capability for an auditing worker.
[DATA_EXPOSURE]: The skill performs read operations on the provided codebase root to detect observability patterns. It writes its findings and reports to a local output directory defined in the contextStore.
[PROMPT_INJECTION]: The instructions contain standard worker directives and mandatory reads of reference files to maintain operational consistency. No bypass or override commands were identified.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect injection as it processes untrusted codebase data and has access to command execution tools. However, it follows a structured two-layer methodology to verify findings, and the execution of analysis tools is restricted to standard auditing workflows.

ln-627-observability-auditor