The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references several external sources for design systems and templates, including public GitHub repositories (e.g., 'tw93/kami', 'op7418/guizang-ppt-skill') and NPM packages ('getdesign'). These references are used for stylistic inspiration and reference material, which is consistent with the skill's primary purpose of providing diverse design recipes.
[COMMAND_EXECUTION]: Some modules within the skill, such as the 'github-dashboard' and 'audio-jingle' sub-skills, utilize the shell capability to gather data from the GitHub API or to dispatch audio generation tasks via platform-specific binaries. These commands are structured templates intended to be used by the agent to fulfill user requests and do not involve arbitrary or high-risk command execution.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests data from external sources like the GitHub API (e.g., issue titles and contributor names) and financial filings. However, the skill explicitly mitigates this risk by instructing the agent to treat such content as untrusted evidence and to ignore any embedded commands or role changes. Per internal guidelines, this surface is classified as low risk.
[CREDENTIALS_UNSAFE]: The skill follows secure practices for credential management, explicitly instructing the agent not to store authentication tokens or private metadata. It also correctly advises the use of standard secret management practices like .env files where necessary.

frontend-design