The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to have the agent review and process external plugin components, including metadata (plugin.toml), documentation (README.md), and source code. This creates an attack surface where malicious instructions embedded in these files could influence the agent's behavior during the review or publishing process.
Ingestion points: Found in references/plugin.md and references/plugin-sharing.md, which instruct the agent to read plugin manifests and documentation.
Boundary markers: No specific boundary markers or instructions to treat external text as untrusted data were identified.
Capability inventory: The skill enables the use of the fba CLI tool, git for repository operations, and ruff for formatting.
Sanitization: The instructions do not mention any sanitization or validation of the input data from plugin files.
[EXTERNAL_DOWNLOADS]: The skill describes a plugin ecosystem where the installer (installer.py) executes uv pip install -r requirements.txt to manage third-party dependencies from external plugin repositories.
[COMMAND_EXECUTION]: The skill provides numerous instructions for executing local CLI tools to manage the project, including fba for formatting and migrations, git for submodule management, and ruff for linting.

fba