The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of potentially untrusted data from Zola project repositories, creating a surface for indirect prompt injection.
Ingestion points: The agent is instructed to read and analyze zola.toml, config.toml, content files, and templates during diagnosis and modification tasks (SKILL.md, workflows/debug-build.md).
Boundary markers: Operations are constrained by specific task-oriented workflows (e.g., workflows/i18n.md, workflows/theme-override.md) which limit the scope of agent actions.
Capability inventory: The skill utilizes the zola CLI to initialize, check, and build sites, which involves local command execution (references/source-registry.md).
Sanitization: The instructions provide explicit guidance on reviewing the | safe filter in Tera templates and emphasize relying on the engine's default auto-escaping (workflows/modify-review.md).
[SAFE]: All external references target official documentation for Zola, W3C accessibility standards, and the Tera template engine, which are well-known and trusted sources.
[SAFE]: The skill encourages security-conscious behavior by recommending isolated output directories for build validation and requiring explicit authorization before applying modifications.

zola