The Agent Skills Directory

[PROMPT_INJECTION]: The skill does not contain instructions that attempt to override agent behavior or bypass safety filters. Instead, it provides defensive guidelines against prompt injection as part of its governance framework.
[DATA_EXFILTRATION]: No commands for accessing sensitive local files or transmitting data to external servers were found. The skill actively discourages passing sensitive information like PII into AI prompts.
[OBFUSCATION]: No obfuscated content, encoded strings, zero-width characters, or hidden text patterns were detected in the instructions or metadata.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download or execution of remote scripts or unverified third-party packages.
[COMMAND_EXECUTION]: The skill includes code snippets for configuration and documentation purposes, but it does not instruct the agent to execute dangerous shell commands or modify system settings.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or other secrets were found. The instructions emphasize the importance of not granting agents access to secrets.
[PRIVILEGE_ESCALATION]: No attempts to acquire elevated permissions or bypass security controls were identified.
[PERSISTENCE_MECHANISMS]: The skill does not contain commands that attempt to establish persistence on the host system.
[METADATA_POISONING]: The metadata correctly reflects the purpose of the skill and does not contain deceptive instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill discusses the risk of indirect prompt injection in the context of LLM01, it does not itself create a vulnerable ingestion surface for untrusted data.

ai-governance