The Agent Skills Directory

[PROMPT_INJECTION]: No instructions attempting to override agent behavior, bypass safety filters, or extract system prompts were detected. The instructions are consistently focused on the stated reliability audit purpose.
[DATA_EXFILTRATION]: The skill does not access sensitive local files (e.g., credentials, tokens) or attempt to exfiltrate data to external servers. It requests specific workload context from the user to perform its analysis.
[REMOTE_CODE_EXECUTION]: No patterns were found for downloading or executing remote code or scripts. All references are to official AWS documentation for the Well-Architected Framework.
[COMMAND_EXECUTION]: The instructions do not involve the execution of shell commands, subprocesses, or any administrative actions on the host system. The skill's operation is purely analytical and reporting-based.
[OBFUSCATION]: No techniques such as Base64 encoding, zero-width characters, or hidden text patterns were used to conceal instructions or URLs.
[INDIRECT_PROMPT_INJECTION]: Although the skill analyzes untrusted data (user codebase), it does not possess exploitable capabilities like network exfiltration or file-writing that could be triggered by malicious instructions embedded in that data.
[EXTERNAL_DOWNLOADS]: The skill references official AWS documentation from a well-known service domain, which is used for context rather than executable content.

reliability-improvement-plan