The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text provided by the user. It contains specific instructions to mitigate indirect prompt injection, explicitly directing the agent to treat the input as material to edit rather than as instructions to follow.
[DATA_EXPOSURE]: No sensitive file paths, hardcoded credentials, or exfiltration patterns were detected. The documentation explicitly warns against handling credentials within the skill context.
[COMMAND_EXECUTION]: No shell commands, subprocess calls, or privilege escalation attempts were found. The skill's 'Boundaries' section strictly limits the agent's capabilities to style editing only.
[REMOTE_CODE_EXECUTION]: No external downloads or remote code execution patterns were identified. All references to external sources are for attribution and documentation purposes.

suede-deslop