Found in the adversarial review of #120.
Capability-matrix tooltips on /mechs/ contradict the new hub and MIM export prose
Where: mechs.md:1019
The capability table renders the CLAW manifest's reasons verbatim, and the PR left that manifest unchanged. The MIM kgx_export tooltip says MIM 'Publishes term mappings rather than a graph ... Adopting KGX would apply here only if it starts emitting nodes and edges'. The PR's new mediaingredientmech.md text says MIM provides 'a standalone KGX ingredient graph' and links a release that 'carries a KGX snapshot'. The CommunityMech sssom_export tooltip says it 'Publishes KGX nodes and edges ... its contribution to kg-microbe is a graph'. The rewritten HUB entry and hub panel say CommunityMech's KGX exporter has published no release and that it releases nothing into the graph.
Failure scenario. A reader hovering the matrix on /mechs/ is told CommunityMech publishes a KGX graph into kg-microbe and MIM emits no graph. The hub panel on the same page, and the MIM page, say the opposite.
Evidence. mechs.md:1019 MIM row: title="kgx_export: disabled. Publishes term mappings rather than a graph; its contribution to kg-microbe is SSSOM. Adopting KGX would apply here only if it starts emitting nodes and edges." mechs.md:1020 CommunityMech row: title="sssom_export: not applicable. Publishes KGX nodes and edges rather than term mappings; its contribution to kg-microbe is a graph...". gh release view mim-supported-2026-09-21 -R CultureBotAI/MediaIngredientMech lists KGX nodes 4,587 and edges 2,112 (mim_nodes.tsv, mim_edges.tsv). gh release list -R CultureBotAI/CommunityMech is empty.
Verifier (confirmed, low). I reproduced every part of the finding. The capability matrix on /mechs/ shows CLAW fleet.yaml reasons word for word. The PR re-pinned CLAW to 2637172, which is still claw main, and left those reasons as they were. Meanwhile it rewrote the HUB entry, the hub panel and communitymech.md to say CommunityMech has published no KGX release. That is true: 0 releases, 0 tags, and the kgx-release workflow has never run. It also added text to mediaingredientmech.md saying MIM ships a standalone KGX graph, with a release carrying 4,587 nodes and 2,112 edges. That is also true. So the same page now says both that CommunityMech "Publishes KGX nodes and edges ... its contribution to kg-microbe is a graph" (tooltip) and that it releases nothing into the graph (hub panel). And the MIM tooltip says MIM publishes "term mappings rather than a graph" and would need KGX only "if it starts emitting nodes and edges", which the MIM page and release show it already has. Before this PR the page agreed with itself: the HUB entry also said CommunityMech had a KGX release. The PR fixed the prose and left the tooltips behind, which is where the contradiction comes from.
Low severity is right. The text appears only in hover tooltips, the table is introduced as "from the manifest", the prose the PR wrote is the accurate side, and the fix belongs in claw's fleet.yaml. It is not a mis-render in this repo. The suggested fix is sound: file an upstream claw issue for the two reasons, since no existing claw issue covers them and claw#358 repeats the stale CommunityMech claim. Until that lands, optionally add a note by the matrix.
Suggested fix. File an upstream claw issue to update the two fleet.yaml reasons, and until it lands add a short note on the page next to the matrix. This belongs with the PR's upstream-defect tracking issue.
Found in the adversarial review of #120.
Capability-matrix tooltips on /mechs/ contradict the new hub and MIM export prose
Where:
mechs.md:1019The capability table renders the CLAW manifest's reasons verbatim, and the PR left that manifest unchanged. The MIM kgx_export tooltip says MIM 'Publishes term mappings rather than a graph ... Adopting KGX would apply here only if it starts emitting nodes and edges'. The PR's new mediaingredientmech.md text says MIM provides 'a standalone KGX ingredient graph' and links a release that 'carries a KGX snapshot'. The CommunityMech sssom_export tooltip says it 'Publishes KGX nodes and edges ... its contribution to kg-microbe is a graph'. The rewritten HUB entry and hub panel say CommunityMech's KGX exporter has published no release and that it releases nothing into the graph.
Failure scenario. A reader hovering the matrix on /mechs/ is told CommunityMech publishes a KGX graph into kg-microbe and MIM emits no graph. The hub panel on the same page, and the MIM page, say the opposite.
Evidence. mechs.md:1019 MIM row: title="kgx_export: disabled. Publishes term mappings rather than a graph; its contribution to kg-microbe is SSSOM. Adopting KGX would apply here only if it starts emitting nodes and edges." mechs.md:1020 CommunityMech row: title="sssom_export: not applicable. Publishes KGX nodes and edges rather than term mappings; its contribution to kg-microbe is a graph...".
gh release view mim-supported-2026-09-21 -R CultureBotAI/MediaIngredientMechlists KGX nodes 4,587 and edges 2,112 (mim_nodes.tsv, mim_edges.tsv).gh release list -R CultureBotAI/CommunityMechis empty.Verifier (confirmed, low). I reproduced every part of the finding. The capability matrix on /mechs/ shows CLAW fleet.yaml reasons word for word. The PR re-pinned CLAW to 2637172, which is still claw main, and left those reasons as they were. Meanwhile it rewrote the HUB entry, the hub panel and communitymech.md to say CommunityMech has published no KGX release. That is true: 0 releases, 0 tags, and the kgx-release workflow has never run. It also added text to mediaingredientmech.md saying MIM ships a standalone KGX graph, with a release carrying 4,587 nodes and 2,112 edges. That is also true. So the same page now says both that CommunityMech "Publishes KGX nodes and edges ... its contribution to kg-microbe is a graph" (tooltip) and that it releases nothing into the graph (hub panel). And the MIM tooltip says MIM publishes "term mappings rather than a graph" and would need KGX only "if it starts emitting nodes and edges", which the MIM page and release show it already has. Before this PR the page agreed with itself: the HUB entry also said CommunityMech had a KGX release. The PR fixed the prose and left the tooltips behind, which is where the contradiction comes from.
Low severity is right. The text appears only in hover tooltips, the table is introduced as "from the manifest", the prose the PR wrote is the accurate side, and the fix belongs in claw's fleet.yaml. It is not a mis-render in this repo. The suggested fix is sound: file an upstream claw issue for the two reasons, since no existing claw issue covers them and claw#358 repeats the stale CommunityMech claim. Until that lands, optionally add a note by the matrix.
Suggested fix. File an upstream claw issue to update the two fleet.yaml reasons, and until it lands add a short note on the page next to the matrix. This belongs with the PR's upstream-defect tracking issue.