The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to transmit user-provided content, including raw text, JSON data, local files, and entire directory structures, to the external domain brewpage.app.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform several system-level operations, including directory compression using zip, generating random strings from /dev/urandom, and executing curl commands to interact with the destination API.
[CREDENTIALS_UNSAFE]: The skill stores ownerToken values, which provide administrative control (deletion/updates) over published content, in a local file at .claude/brewpage-history.md. While the skill includes instructions to keep this file private, it represents local storage of sensitive credentials.
[DYNAMIC_EXECUTION]: The skill dynamically generates a ZIP archive of local directories at runtime using the zip utility and creates temporary files via mktemp to facilitate the upload process.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the user's environment (files, directories, and text) which could contain malicious instructions. While the skill uses jq to safely wrap some content types, it zips and uploads directory contents without inspection.
Ingestion points: The content_arg parameter in SKILL.md, which can point to any local file, directory, or raw text string.
Boundary markers: Absent; there are no specific markers or warnings to the agent to ignore instructions embedded within the content being published.
Capability inventory: The Bash tool is used for file system reads, directory zipping, and network transmission via curl across all publishing blocks in SKILL.md.
Sanitization: The skill uses jq to encode text and JSON content into API payloads, but no sanitization or filtering is performed on the contents of files or directories beyond excluding specific sensitive filenames (e.g., .env, .git).

brewpage-publish