No user-submitted related servers found.
Scored across 51 tools
Tools are grouped into clearly named families (gates_*, state_*, worker_*, loop_*, project_docs_*, self_verify_*), and within each family the action differs (init/lint/check/status/report/abandon; read/write/list/prune/doctor/maintain). A few boundaries are still soft — command_policy_check vs command_policy_audit vs command_fake_run all 'evaluate policy without executing', and gates_check vs gates_status vs gates_lint require reading the description carefully to separate. The `issueops_execution` tool bundles seven lease actions into one entry, which makes it a poor neighbor to the single-action tools.
Most tools follow snake_case with a domain prefix plus verb (gates_check, state_read, worker_enqueue, project_docs_revise), which is readable and predictable. However, several tools are noun-only or verb-only (commit_policy, contract_schema, skill_manifest, docs_index, self_verify, self_augment), and issueops_execution packs multiple verbs behind an `action` parameter. The mix is legible but not a uniform verb_noun convention.
51 tools is well past the point where an agent can hold the surface in mind, and several clusters are only marginally distinct. While the harness genuinely spans many domains, the set would be stronger as a smaller number of well-scoped tools or split into separate servers. Individual tools earn their place, but the aggregate is heavy.
Coverage across the harness domain is broad: state lifecycle (read/write/list/prune/doctor/maintain), worker lifecycle (enqueue/list/status/cancel/run), gates (init/lint/check/status/report/abandon), loops, docs routing/read/revise/append, and self-verification history/promotion. Minor gaps remain — e.g. no channel discovery/creation tool alongside channel_send/recv, and no worker deletion or bulk cleanup — but these are workable around.
