The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes official integrations and libraries from trusted organizations.
It employs the anthropics/claude-code-base-action, an official action from a trusted vendor, to handle content generation.
The workflow dynamically installs the @google/genai package from the official NPM registry, which is a well-known library from a trusted organization.
[COMMAND_EXECUTION]: The workflow executes shell commands for repository management and automated script execution.
The AI agent's tool access is restricted to a specific whitelist of safe commands (git, ls, date), adhering to the principle of least privilege and preventing arbitrary command execution.
[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection because it processes data from external sources to drive AI behavior.
Ingestion points: The workflow reads data from content/topics.yml and processes user-provided strings via the topic input in workflow_dispatch events.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included when interpolating topics into the generation prompt.
Capability inventory: The agent has permissions to write files to the repository and create pull requests using the GITHUB_TOKEN. It also has restricted access to system tools through a controlled bash environment.
Sanitization: There is no evidence of sanitization or structural validation performed on the ingested topic strings before they are sent to the LLM.

blog-generator