The Agent Skills Directory

[SAFE]: The skill consists of markdown instructions and YAML workflow definitions designed to guide an AI agent through software development and auditing tasks. No code or prompt-based security threats were identified.
[COMMAND_EXECUTION]: Several workflows permit the use of the Bash tool to interact with standard development utilities such as git, terraform, and npm. These are consistent with the skill's primary purpose as a developer assistant and are governed by the platform's tool configuration.
[EXTERNAL_DOWNLOADS]: The deep-research and gather-review components utilize network-enabled tools (WebFetch, WebSearch, gh) to retrieve relevant project information or public data. The instructions include explicit safety constraints, such as prohibiting the download of executable files and avoiding suspicious domains.
[PROMPT_INJECTION]: While the skill ingests external data like web search results and execution logs, the instructions incorporate boundary markers and specific verification steps (e.g., in INSTRUCTION_STYLE_GUIDE.md) to prevent obedience to embedded instructions within untrusted content.
[CREDENTIALS_UNSAFE]: Sample configuration files include placeholders for API keys (e.g., sk-ant-...), but no actual secrets are hardcoded. The skill explicitly instructs agents never to write passwords or tokens into the code.

j5ik2o-takt-analyzer