The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is primarily designed to upload local files and recursive directory structures to the external service brewpage.app using curl. This facilitates the transfer of local data to a public hosting environment.- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform file system operations such as cat for reading file contents, zip for compressing directories, and file --mime-type for identifying file types, followed by curl for network transmission.- [CREDENTIALS_UNSAFE]: The skill generates and stores ownerToken values—API keys that grant permission to delete or update published content—in a local file at .claude/brewpage-history.md. These tokens are stored in plaintext on the local filesystem.- [INDIRECT_PROMPT_INJECTION]: The skill resolves which files or directories to publish based on free-form natural language prompts. This represents a potential vulnerability if the agent processes untrusted data containing instructions to upload sensitive paths.
Ingestion points: The $ARGUMENTS field in SKILL.md accepts free-form text used to determine content to publish.
Boundary markers: The workflow includes AskUserQuestion steps for namespace selection and password confirmation before the upload is executed.
Capability inventory: The skill uses Read, Bash, and Glob tools for file access and Bash for network requests.
Sanitization: The skill uses jq to encode content into JSON payloads, providing a layer of protection against payload manipulation.

publish