The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no attempts to bypass safety filters or override system instructions. All commands are project-specific quality constraints.
[DATA_EXFILTRATION]: No network operations, credential harvesting, or access to sensitive local files (like .env or .ssh) were identified.
[REMOTE_CODE_EXECUTION]: There are no patterns involving external downloads (curl|bash), package installations, or dynamic execution of untrusted code.
[COMMAND_EXECUTION]: The instructions do not invoke any dangerous shell commands or attempt to acquire elevated privileges (sudo/chmod).
[OBFUSCATION]: The content is written in clear markdown with no hidden characters, Base64 encoding, or homoglyph attacks detected.
[INDIRECT_PROMPT_INJECTION]: While the skill monitors user-initiated code changes, it does not ingest untrusted external data (like web content or emails) that could be used for injection attacks.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use the shell execution placeholder syntax (!command) in its SKILL.md definition.

preflight