Install any skill in seconds. Free to start, no credit card required.
Get Started Free →razzant/claudexor
Use the local Claudexor control plane for harness-agnostic coding work across Claude Code, Codex, Cursor, and OpenCode.
Passed 17 of 20 test cases with the skill, and 10 of 20 without it.
and 5 more in the full report
8 fixed12 passed either way3 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.8-flash on 10/21/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.
The `claudexor_apply_check` tool returned `applyEligibility.eligible = true` for a completed run patch, and the user explicitly requested patch delivery. Does the MCP tool itself modify the local files, or how is delivery performed?
| Checked | Without skill | With skill |
|---|---|---|
| Explains that patch application is executed by the host invoking the Claudexor CLI apply path. | Missed | Passed |
The instructions your model reads when the skill loads.
Use the MCP tools supplied by this plugin. They connect to the user's local Claudexor installation; they do not provide accounts or remote model access.
claudexor command. If the
MCP server is missing, ask the user to run npm install -g claudexor, then
restart the host. Do not download a runtime implicitly.Start read-only unless the user explicitly asked to change or create files.
claudexor_status to inspect the aggregate/default-store projection,
but do not use it as a universal admission veto. An unpinned request may be
admitted by the canonical account pool through a ready exact profile even
when default doctor is unavailable; only a genuinely profile-less/default
fallback requires doctor status ok and the requested intent. Aggregate
status neither proves nor vetoes a pooled or explicitly named profile.claudexor_accounts before choosing an account or reviewer identity.
It is the read-only Accounts view of registered profiles, readiness, quota
freshness, and the daemon's next_up routing projection; the default call
combines current quota and registry facts with retained readiness observations.
One coalesced cold acquisition, including failure, is followed by passive reads
with the original check time; viewing does not renew old evidence.
fresh: true requests the atomic snapshot, an expensive explicit refresh
to use only when acting on staleness matters. Check the selected row's age:
retained available/passed describes its last observation, while actual
execution performs its own current admission checks. unknown,
not_run, or stale quota means uncertain, not absent. Never substitute
aggregate doctor status, another profile's probe, or a host/default login.
Never initiate login or OAuth merely because a row is unknown; ask the user
for explicit authority. An explicit reviewer credentialProfileId is
strict and never falls back, while an omitted id uses the canonical accounttest cases passed, vs 10 / 20 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.