experiment-queue

SSH job queue for multi-seed/multi-config ML experiments with OOM-aware retry, stale-screen cleanup, and wave-transition race prevention. Use when user says "batch experiments", "队列实验", "run grid", "multi-seed sweep", "auto-chain experiments", or when /run-experiment is insufficient for 10+ jobs that need orchestration.

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Quality

Content

77%

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

The content is highly actionable with a well-validated, clearly sequenced workflow and concrete executable scripts. Its main weaknesses are verbosity from audit/rationale and version-history narrative, and progressive-disclosure gaps where detailed content stays inline and several referenced shared-references files are missing from the bundle.

Suggestions

Trim meta-narrative that doesn't help execution — the "Rationale / Source" post-mortem paragraph (incl. "Codex GPT-5.5 xhigh"), the "Why This Exists" audit story, and the "Phase 3.3 move, Arch C" / follow-up-tracking asides — to tighten conciseness toward the level-3 bar.

Resolve or remove the dead shared-references links (../shared-references/external-cadence.md and ../shared-references/compute-env-contract.md); if those files aren't bundled, inline the essential guidance they were meant to carry so the skill is self-contained.

Consider moving the long bash resolution-chain fallback (Step 3a), the comparison table, and the Example Session into a separate reference file, leaving SKILL.md as a lean overview that points one level deep — improving progressive_disclosure.

Dimension	Reasoning	Score
Conciseness	The body is mostly operational and useful, but carries padding that does not aid execution — the "Rationale / Source" post-mortem notes ("Codex GPT-5.5 xhigh"), the "Why This Exists" audit narrative, version-history commentary ("Phase 3.3 move, Arch C"), and a follow-up tracking note ("Tracking /monitor-experiment queue-state integration as a follow-up"). Not the level-3 'every token earns its place' bar.	2 / 3
Actionability	Provides fully executable, copy-paste-ready bash launch/resume commands, concrete YAML manifests, an OOM regex, and jq monitoring snippets with real variable handling — matching the 'fully executable code/commands; copy-paste ready' anchor.	3 / 3
Workflow Clarity	The 5-step workflow is clearly sequenced with an explicit validation checkpoint (Step 2 Pre-flight checks SSH/conda/cwd/preconditions/GPUs and reports blocked jobs), feedback loops (OOM retry, resume-on-restart), and output-file-verified completion for batch operations.	3 / 3
Progressive Disclosure	Section organization exists and the two scripts/ helpers are real bundle files referenced by path, but the body is a long monolithic document with inline content that could be split out, and two prominent shared-references links (external-cadence.md, compute-env-contract.md) point to files absent from the bundle — references present but not all reliably signaled/available.	2 / 3
	Total	10 / 12 Passed

Description

100%

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

The description is specific, trigger-rich, and complete, covering both capabilities and explicit use-conditions in a concise third-person voice with minimal fluff. It also cleanly distinguishes itself from the related /run-experiment skill.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities — "OOM-aware retry, stale-screen cleanup, and wave-transition race prevention" for multi-seed/multi-config ML experiments on SSH — rather than vague abstractions, matching the 'multiple specific concrete actions' anchor.	3 / 3
Completeness	Explicitly answers both what ("SSH job queue for multi-seed/multi-config ML experiments with OOM-aware retry...") and when ("Use when user says... or when /run-experiment is insufficient for 10+ jobs"), matching the 'clearly answers both what AND when with explicit triggers' anchor.	3 / 3
Trigger Term Quality	Provides natural user phrases a person would actually say — "batch experiments", "队列实验", "run grid", "multi-seed sweep", "auto-chain experiments" — giving broad coverage including a multilingual term, matching the 'good coverage of natural terms' anchor.	3 / 3
Distinctiveness Conflict Risk	Occupies a clear niche (SSH GPU-server job orchestration) with distinct triggers and an explicit disambiguator against /run-experiment, making it unlikely to fire for the wrong skill.	3 / 3
	Total	12 / 12 Passed

Validation

81%

Warnings & errors only

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

Validation — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 1 suspicious	Warning

	Total	13 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Commit: 82076e5

Reviewed: 5 days ago

Table of Contents

Discovery Implementation Validation

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.