Menu
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/ben-manes/caffeine --skill audit-lifecycleThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Regenerate the window climber's adversarial trap traces and run the behavioral gate against the real cache in the simulator
Adversarial workload search for eviction-policy regret. Crafts, shrinks, and classifies workloads on which the adaptive window climber fails to close the gap to its achievable optimum, and routes each failure class to the mechanism responsible (controller, policy structure, or recovery layer)
Price each algorithmic step of the window climber by removing it, to find steps that no longer earn their keep and branches that no longer fire
Showing SKILL.md
Based on SOC occupation classification
| name | audit-lifecycle |
| description | Analyze shutdown, close, and GC-of-cache lifecycle correctness |
| context | fork |
| agent | auditor |
| disable-model-invocation | true |
Analyze the cache for defects during shutdown, close, and garbage collection of the cache instance itself.
Assume at least one shutdown or lifecycle race exists. If your analysis yields zero findings, re-examine scheduler/executor lifecycle and in-flight operations during close — explain specifically why no race is possible.
Operations racing with close/cleanup: get/put/compute during cleanUp() or GC? close() state transition? Operations after close()?
Scheduler lifecycle: Does the scheduled future prevent cache GC? Can scheduler fire after cache is collected? Is the task cancelled on close?
Executor lifecycle: Reference cycle between cache and executor? RejectedExecutionException handling when executor is shut down?
Removal listener drain on shutdown: Are pending notifications delivered or dropped? Notifications lost if executor shuts down first?
In-flight operations during shutdown:
Multiple close calls: Is close() idempotent? Concurrent close() corruption?
Finalizer / Cleaner interactions: Reference queues outliving the cache? Reference queue processing accessing dead cache?
jcache manager/cache close: Is teardown symmetric with setup — the OSGi context-classloader swap applied on destroy/close as on create/get; in-flight futures awaited; user Closeables (writer, loader, listeners, expiry) closed exactly once under the right classloader?
For each defect: state the lifecycle transition, provide a concrete scenario, state the observable incorrect behavior.