The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of external software and extensions.
Instructs the user to install the markit-ai package globally using bun add -g markit-ai.
The plugin system supports downloading and installing extensions from external sources, including NPM and arbitrary GitHub repositories (e.g., markit plugin install git:github.com/user/markit-plugin-ocr).
[DYNAMIC_EXECUTION]: The skill employs a plugin architecture that allows for runtime code execution.
Plugins can be loaded from NPM, Git, or local file paths (e.g., ./my-plugin.ts).
These plugins are capable of overriding existing converters or adding new ones, allowing third-party code to intercept and process data during conversion tasks.
[COMMAND_EXECUTION]: The skill is centered around the execution of the markit CLI tool and its various subcommands.
It documents multiple shell commands for file conversion, configuration management, and environment setup.
The markit onboard command is designed to modify project files such as CLAUDE.md or AGENTS.md to persist instructions for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection due to its primary function.
Ingestion points: The tool processes data from untrusted external sources including URLs, PDFs, spreadsheets, RSS feeds, and GitHub repositories.
Boundary markers: The skill documentation does not specify the use of delimiters or 'ignore' instructions when processing this external content.
Capability inventory: The tool can write to the local file system (-o, -i), install plugins, and send data to AI providers (OpenAI, Anthropic) for transcription or description.
Sanitization: There is no mention of sanitization or filtering for the content converted from external documents or webpages before it is returned to the agent's context.

jb-markit