The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions focus on diagnosing context issues and do not contain directives to bypass safety filters or override system instructions. Benign usage of terms like 'IMPORTANT' within code comments was identified and confirmed to be non-malicious.
[DATA_EXFILTRATION]: No network access or sensitive file reading operations are present in the scripts or documentation. The code performs purely local string processing.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or private tokens were found in the skill files. Standard best practices for secret management are not applicable here as no secrets are required.
[REMOTE_CODE_EXECUTION]: The Python script contains local logic for text analysis and does not download or execute external code from remote sources.
[OBFUSCATION]: No base64, hex encoding, zero-width characters, or homoglyph attacks were identified in the content. All strings and logic are presented in plain text.
[DYNAMIC_EXECUTION]: The analysis logic uses standard Python string and regex operations; there is no use of eval(), exec(), or dynamic module loading for runtime code assembly.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external context data for diagnosis, it lacks any high-privilege capabilities (such as file system writes, network requests, or shell command execution) that could be exploited through data-embedded instructions.

context-degradation