DOCS+PLG: skill updates, PIHOLEMON corrected source

.claude/skills/plugin-readme/SKILL.md

@@ -41,7 +41,7 @@ Before concluding a plugin has no attribution to record, grep its script for a c
 - `TBC` or similarly empty content, especially for a prominent feature.
 - Duplicate or orphaned sections (e.g. two `### Usage` headings) - usually a merge/edit artifact.
 - Sibling non-README files (a provider-specific sub-guide, a translated `README_<LANG>.md`) that aren't linked from the plugin's own `README.md` - `docs/gen_plugin_pages.py` generates a page for every `*.md` in the plugin folder, but only reachable if something links to it.
-- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates this, but the docs site (`mkdocs`, Python-Markdown) terminates the list right there - the table and everything after it fall out as orphaned paragraphs, and the *next* bullet renders as a literal `-`-prefixed line of text instead of a list item. Looks fine in the PR diff on GitHub, breaks only once published. Fix: de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
+- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates loose (2-3 space) indentation; the docs site (`mkdocs`, Python-Markdown) only keeps the block nested inside that list item at a full 4-space indent - anything less and the block falls out as an orphaned, unindented paragraph/table right after the list closes. Separately, and regardless of indent width: a bullet that follows the block *without* a blank line in between merges into it as plain text instead of parsing as a new list item - confirmed this still breaks even at the correct 4-space indent, so fixing the indent alone isn't sufficient. Looks fine in the PR diff on GitHub, breaks only once published. Simplest, most portable fix (works regardless of either rule): de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
 
 ## Reference
 

.gemini/skills/plugin-readme/plugin-readme-skill.md

@@ -41,7 +41,7 @@ Before concluding a plugin has no attribution to record, grep its script for a c
 - `TBC` or similarly empty content, especially for a prominent feature.
 - Duplicate or orphaned sections (e.g. two `### Usage` headings) - usually a merge/edit artifact.
 - Sibling non-README files (a provider-specific sub-guide, a translated `README_<LANG>.md`) that aren't linked from the plugin's own `README.md` - `docs/gen_plugin_pages.py` generates a page for every `*.md` in the plugin folder, but only reachable if something links to it.
-- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates this, but the docs site (`mkdocs`, Python-Markdown) terminates the list right there - the table and everything after it fall out as orphaned paragraphs, and the *next* bullet renders as a literal `-`-prefixed line of text instead of a list item. Looks fine in the PR diff on GitHub, breaks only once published. Fix: de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
+- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates loose (2-3 space) indentation; the docs site (`mkdocs`, Python-Markdown) only keeps the block nested inside that list item at a full 4-space indent - anything less and the block falls out as an orphaned, unindented paragraph/table right after the list closes. Separately, and regardless of indent width: a bullet that follows the block *without* a blank line in between merges into it as plain text instead of parsing as a new list item - confirmed this still breaks even at the correct 4-space indent, so fixing the indent alone isn't sufficient. Looks fine in the PR diff on GitHub, breaks only once published. Simplest, most portable fix (works regardless of either rule): de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
 
 ## Reference
 

.github/skills/plugin-readme/SKILL.md

@@ -41,7 +41,7 @@ Before concluding a plugin has no attribution to record, grep its script for a c
 - `TBC` or similarly empty content, especially for a prominent feature.
 - Duplicate or orphaned sections (e.g. two `### Usage` headings) - usually a merge/edit artifact.
 - Sibling non-README files (a provider-specific sub-guide, a translated `README_<LANG>.md`) that aren't linked from the plugin's own `README.md` - `docs/gen_plugin_pages.py` generates a page for every `*.md` in the plugin folder, but only reachable if something links to it.
-- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates this, but the docs site (`mkdocs`, Python-Markdown) terminates the list right there - the table and everything after it fall out as orphaned paragraphs, and the *next* bullet renders as a literal `-`-prefixed line of text instead of a list item. Looks fine in the PR diff on GitHub, breaks only once published. Fix: de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
+- A table (or any multi-line block) indented under a bullet as that list item's continuation. GitHub's renderer tolerates loose (2-3 space) indentation; the docs site (`mkdocs`, Python-Markdown) only keeps the block nested inside that list item at a full 4-space indent - anything less and the block falls out as an orphaned, unindented paragraph/table right after the list closes. Separately, and regardless of indent width: a bullet that follows the block *without* a blank line in between merges into it as plain text instead of parsing as a new list item - confirmed this still breaks even at the correct 4-space indent, so fixing the indent alone isn't sufficient. Looks fine in the PR diff on GitHub, breaks only once published. Simplest, most portable fix (works regardless of either rule): de-nest it - end the bullet's text, blank line, then the table/block as top-level (unindented) content, blank line, then resume the list as a fresh block.
 
 ## Reference
 

docs/PLUGINS_OVERVIEW.md

@@ -58,6 +58,7 @@ The **Plugin docs** links below open each plugin's README rendered as part of th
 | `DHCPLSS`       | [dhcp_leases](plugins/dhcp_leases.md)                         | 📥/🆎   | Import devices from DHCP leases           |          |          |
 | `DHCPSRVS`      | [dhcp_servers](plugins/dhcp_servers.md)                       | ♻        | DHCP servers                              |          |          |
 | `DIGSCAN`       | [dig_scan](plugins/dig_scan.md)                               | 🆎       | Dig (DNS) Name resolution                 |          |          |
+| `DOCKERDISC`    | [dockerdisc](plugins/dockerdisc.md)                           | ♻        | Enriches known Docker hosts with their running containers |          |          |
 | `FREEBOX`       | [freebox](plugins/freebox.md)                                  |📥/♻/🆎  | Pull data and names from Freebox/Iliadbox |          |          |
 | `FRITZBOX`      | [fritzbox](plugins/fritzbox.md)                                | 📥       | Fritz!Box device scanner via TR-064       |          |          |
 | `ICMP`          | [icmp_scan](plugins/icmp_scan.md)                             | ♻        | ICMP (ping) status checker                |          |          |

server/plugins/pihole_monitor/config.json

@@ -524,7 +524,6 @@
     },
     {
       "column": "extra",
-      "mapped_to_column": "scanSourcePlugin",
       "css_classes": "col-sm-3",
       "show": true,
       "type": "label",
@@ -533,6 +532,20 @@
       "localized": ["name"],
       "name": [{ "language_code": "en_us", "string": "Detail" }]
     },
+    {
+      "column": "Dummy",
+      "mapped_to_column": "scanSourcePlugin",
+      "mapped_to_column_data": {
+        "value": "PIHOLEMON"
+      },
+      "css_classes": "col-sm-2",
+      "show": false,
+      "type": "none",
+      "default_value": "",
+      "options": [],
+      "localized": ["name"],
+      "name": [{ "language_code": "en_us", "string": "N/A" }]
+    },
     {
       "column": "dateTimeChanged",
       "css_classes": "col-sm-2",

test/plugins/test_adguard_import.py

@@ -17,9 +17,11 @@ is a permanent binding, not evidence of current connectivity), passed via
 helpVal4 (config.json maps helpVal4 -> scanPresence).
 """
 
+import importlib.util
 import os
 import sys
 import types
+from pathlib import Path
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -47,8 +49,6 @@ def _load_adguard_import_module():
     stub("conf", tz=None)
     stub("pytz", timezone=MagicMock(return_value="UTC"))
 
-    import importlib.util
-    from pathlib import Path
     module_path = Path(__file__).resolve().parents[2] / "server" / "plugins" / "adguard_import" / "adguard_import.py"
     spec = importlib.util.spec_from_file_location("adguard_import_script", module_path)
     module = importlib.util.module_from_spec(spec)

test/plugins/test_plugin_conventions.py

@@ -415,8 +415,16 @@ def test_scan_source_plugin_matches_unique_prefix(plugin_name):
         if col.get('mapped_to_column') != 'scanSourcePlugin':
             continue
         value = (col.get('mapped_to_column_data') or {}).get('value')
-        if value is None:
-            continue
+        assert value is not None, (
+            f"{plugin_name}: column {col.get('column')!r} maps to scanSourcePlugin "
+            f"but has no static \"mapped_to_column_data\": {{\"value\": ...}} - scanSourcePlugin "
+            f"identifies which plugin produced a CurrentScan row and must be a single static "
+            f"value (its own unique_prefix, {prefix!r}), not a per-row value from another "
+            f"field. Mapping a per-row field here (e.g. a free-text detail column) writes "
+            f"that field's actual value into scanSourcePlugin instead, breaking "
+            f"update_devices_data_from_scan()'s per-plugin grouping the same way a wrong "
+            f"static value does."
+        )
         assert value == prefix, (
             f"{plugin_name}: scanSourcePlugin's static value is {value!r}, but "
             f"unique_prefix is {prefix!r}. These must match exactly - "