> ## Documentation Index
> Fetch the complete documentation index at: https://partners.trybloom.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Use the downloaded Skill

> Store and mount the generated Brand Skill for an agent.

The downloaded ZIP is a ready-to-mount Agent Skill. Extract and store the whole
folder without flattening or renaming its contents.

A completed Skill has this shape:

```text theme={null}
SKILL.md
references/**/*.md    # task-specific guidance, when present
assets/images/*       # selected images, when present
assets/fonts/*        # website fonts, when available
```

`SKILL.md` is the entrypoint. It contains routing instructions that tell an
agent which files under `references/` are relevant to the current task. Image
and font paths are relative to the Skill folder.

## If your platform supports Agent Skills

Install or mount the extracted folder as one Skill. Keep `SKILL.md` at the root
and make the sibling `references/` and `assets/` directories available through
the same folder.

The agent should begin with `SKILL.md`, then read the references routed for the
task it is performing.

## If you run a custom agent

Use `SKILL.md` as the initial instruction file and give the agent read access to
the complete extracted directory. Resolve every referenced Markdown file,
image, and font relative to that directory.

Do not concatenate every Markdown file into every prompt. That wastes context
and bypasses the routing already encoded in `SKILL.md`.

## Preserve the folder as one unit

* Store every file from the ZIP, including assets not needed by the first task.
* Preserve filenames and relative paths.
* Keep the Skill scoped to the brand that produced it.
* Treat each downloaded folder as one immutable version. Swap the whole folder
  when updating instead of mixing files from different versions.

The agent can use the stored Skill independently after download. It does not
need to call Bloom while performing later brand tasks.