docs: mark custom prompts and tools as experimental and unsupported

Add disclaimers to prompts, tools, and configuration documentation
noting that custom prompt and tool configuration is experimental.
Bugs with the ingestor, reviewer, and core tooling should still be
filed, but custom prompt debugging and token consumption issues
from custom configurations are not supported.

Assisted-by: claude-opus-4.6
Signed-off-by: derekbarbosa <derekasobrab@gmail.com>

Author: derekbarbosa

README.md

@@ -91,7 +91,7 @@ cd sashiko
 Copy an example config to get started. For a full reference of every
 setting, see the [Configuration Reference](docs/configuration.md).
 
-Optional customization sections:
+Optional customization sections (experimental -- see linked docs for support policy):
 *   **Tools**: Configure which AI tools are enabled (optional). See [docs/tools.md](docs/tools.md) for details.
 *   **Prompts**: Customize review stages and prompts (optional). See [docs/prompts.md](docs/prompts.md) for details.
 

docs/configuration.md

@@ -154,7 +154,9 @@ Optional array of additional git remotes to track.
 | `max_total_tokens` | integer | `5000000` | Maximum cumulative uncached tokens (input + output) per review. Cached tokens are excluded. Set to 0 to disable. |
 | `max_total_output_tokens` | integer | `500000` | Maximum cumulative output tokens per review. Set to 0 to disable. |
 
-### `[tools]` (optional)
+### `[tools]` (optional, experimental)
+
+> **Experimental / Unsupported** -- see [AI Tools](tools.md) for details.
 
 Configure which AI tools are available during review. See [AI Tools](tools.md) for full documentation.
 
@@ -164,7 +166,9 @@ Configure which AI tools are available during review. See [AI Tools](tools.md) f
 | `disabled` | list | `[]` | Denylist of tool names. Takes precedence over `enabled`. |
 | `custom` | list | `[]` | Custom shell-based tool definitions. |
 
-### `[prompts]` (optional)
+### `[prompts]` (optional, experimental)
+
+> **Experimental / Unsupported** -- see [Prompt Customization](prompts.md) for details.
 
 Customize review prompts and stages. See [Prompt Customization](prompts.md) for full documentation.
 

docs/prompts.md

@@ -1,5 +1,11 @@
 # Prompt Customization
 
+> **Experimental / Unsupported:** Custom prompt configuration is experimental
+> and provided as-is. Bugs and issues with the Sashiko ingestor, reviewer, or
+> core tooling can and should still be filed. However, we will not provide
+> support for debugging custom prompts or troubleshooting token consumption
+> caused by custom prompt configurations.
+
 Sashiko's multi-stage review pipeline loads its prompts from a
 configurable directory. You can override prompt content, reorder or
 disable stages, define template variables, and point to a remote prompt

docs/tools.md

@@ -1,5 +1,11 @@
 # AI Tools
 
+> **Experimental / Unsupported:** Custom tool configuration is experimental
+> and provided as-is. Bugs and issues with the Sashiko ingestor, reviewer, or
+> core tooling can and should still be filed. However, we will not provide
+> support for debugging custom tool definitions or troubleshooting token
+> consumption caused by custom tool configurations.
+
 Sashiko gives the AI reviewer a set of built-in tools for navigating the
 codebase under review. All built-in tools are enabled by default; you can
 restrict or extend them via `Settings.toml`.