CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

cmux-connector

Connect to already-running cmux panes on the remote Mac (built by cmux-fanout) and inject caller-supplied work into them from Windows — one runspace per pane, fire-and-forget or run-and-wait, broadcast or per-pane commands, ShellManaged/CallerManaged/RawInteractive profiles. Use AFTER cmux-fanout has built a grid, when you want to run/administer commands across those existing panes concurrently or steer agent REPLs in them. Requires PowerShell 7.1+, key-auth SSH to the Mac, and a pane map written by cmux-fanout. NOT for building panes (use cmux-fanout), local Windows grids (use psmux-connector), or in-process parallelism (use the Task tool).

Quality

95%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

96%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

An exceptionally tight operational skill: executable commands covering every common invocation pattern, explicit pre-send validation and error-recovery loops for a batch operation, and dense prose with no padding. The single structural improvement would be moving the parameter reference and identity-validation internals into a bundled reference file so SKILL.md stays a pure overview.

DimensionReasoningScore

Conciseness

Every section is operationally dense with zero padding: nothing explains concepts Claude already knows — no SSH tutorial, no PowerShell basics — and each line carries skill-specific facts ("every remote touch a ~258 ms SSH call", "-MaxTypedChars, default 4096", "Timeout ⇒ timedout-unknown — the remote command may still be running; it is NEVER auto-retried"). This matches the level-5 anchor (lean, assumes competence, every token earns its place); level 4 would require identifiable over-explanation to trim, and none is evident.

5 / 5

Actionability

Five ready-to-run invocations in 'How to run' cover the common cases — broadcast ("& $connect -WorkspaceName fanout-... -Command 'git pull; npm test' -Wait"), per-pane (-Commands array), REPL steering (-Profile RawInteractive), subset (-PaneIndex (0..3)), and sentinel-driven agent tasks — plus executable Pester and smoke-test commands. Placeholders like "<...>\__solutions\" are environment-specific values equivalent to the anchor's "file.pdf", so this fits the level-5 anchor (copy-paste-ready commands covering common cases) rather than level 4's 'minor gaps'.

5 / 5

Workflow Clarity

The lifecycle is explicitly sequenced ("cmux-fanout (build grid + map) → cmux-connector (inject / administer) → cmux-fanout teardown.ps1"), and this batch operation has thorough validation: pre-send identity validation ("the live tree must contain every manifest surface with NO extras"), fail-before-send on unknown indices, and a documented status taxonomy with failure modes ("failed | timedout-unknown | ..."). Explicit feedback loops are present — the context-limit recovery loop ("send /clear, wait for -ReadyPattern ... re-send ONCE") and the CMUX_SHAPE drift procedure ("run fanout's tests\smoke\03-discovery.ps1 and fix ONLY the key list") — matching the level-5 anchor; the batch-operation cap of 3 does not apply since validation is extensive.

5 / 5

Progressive Disclosure

No bundle reference files exist (no references/, scripts/, or assets/ directories), and the body is well sectioned (Prerequisites, How to run, profiles table, Key parameters, Results, gotchas, Tests, Cleanup) with clearly signaled one-level references ("../cmux-fanout/setup.md", the smoke-test paths). It falls at level 4 rather than 5 because reference-grade detail — the full parameter semantics in 'Key parameters' and the identity-validation internals — is inlined in SKILL.md where a skill of this complexity would ideally split it into a separate reference file, leaving only an overview plus pointers.

4 / 5

Total

19

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: it states concrete capabilities in third person, gives an explicit 'Use AFTER cmux-fanout...' trigger clause, lists hard requirements, and actively redirects the three most likely wrong-skill invocations. The only weakness is that a handful of natural user phrasings for triggering are implied rather than enumerated, capping trigger-term coverage just short of comprehensive.

DimensionReasoningScore

Specificity

Multiple concrete actions are enumerated: "Connect to already-running cmux panes", "inject caller-supplied work into them from Windows", "one runspace per pane, fire-and-forget or run-and-wait, broadcast or per-pane commands, ShellManaged/CallerManaged/RawInteractive profiles" — comprehensive coverage of the skill's capabilities, matching the level-5 anchor (multiple specific concrete actions). Not 4: unlike the "minor gaps in coverage" anchor, the broadcast/per-pane, wait/no-wait, and all three profile modes are explicitly named.

5 / 5

Completeness

Explicitly answers both: WHAT — "Connect to already-running cmux panes ... and inject caller-supplied work into them from Windows"; WHEN — "Use AFTER cmux-fanout has built a grid, when you want to run/administer commands across those existing panes concurrently or steer agent REPLs in them", plus requirements and an explicit NOT-for clause. This matches the level-5 anchor (clear, explicit what AND when with concrete trigger phrases); level 4 would require a weaker or less explicit 'when', which is not the case here.

5 / 5

Trigger Term Quality

Good natural keyword coverage: "run/administer commands across those existing panes concurrently", "steer agent REPLs", "inject caller-supplied work", "fire-and-forget or run-and-wait", plus redirect terms like "cmux-fanout", "psmux-connector", and "Task tool". A few natural phrasings a user might say (e.g., "send work to the panes", "check/drive the agents on the Mac") are only implied, so it sits at the level-4 anchor (good coverage, a few natural terms missing) rather than the comprehensive-synonym level-5 anchor.

4 / 5

Distinctiveness Conflict Risk

Clear niche with explicit conflict-avoidance redirects: "NOT for building panes (use cmux-fanout), local Windows grids (use psmux-connector), or in-process parallelism (use the Task tool)", and "Use AFTER cmux-fanout has built a grid". This is stronger than the level-4 anchor ("mostly distinct; minor overlap risk") — the sibling-skill boundaries are actively disambiguated rather than merely distinct.

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

KennethJefferson/cmux_test

Path

__solutions/cmux-connector/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.