The Agent Skills Directory

[PROMPT_INJECTION]: The instructions do not contain patterns designed to bypass safety filters or override system constraints. It includes clear restrictive rules for the agent, such as the prohibition of external posting.- [DATA_EXPOSURE_AND_EXFILTRATION]: No hardcoded credentials, sensitive system paths, or unauthorized network calls were detected. The skill explicitly states that it must never post or send content externally, requiring manual human delivery.- [OBFUSCATION]: No Base64, zero-width characters, homoglyphs, or hidden URLs were found in the skill metadata or body. References to 'hex' are limited to standard color palette definitions.- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: There are no package installations (pip, npm) or remote code download patterns. The skill relies entirely on text generation within the agent environment.- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data (signal feeds/news), but the risk is mitigated by the lack of automated capabilities (no file-write or network-send tools) and the requirement for a human to approve the final output.- [DYNAMIC_EXECUTION]: No usage of unsafe deserialization, runtime compilation, or shell execution based on user-supplied strings was found.

linkedin-ghostwriting