Analyze Shark agent timing logs to recommend SHARK_LOOP_TIMEOUT and SHARK_MAX_LOOPS values.
npx skills add https://github.com/keugenek/shark --skill shark-autotune
Please help me install this Agent Skill. Skill: shark-autotune Source: https://github.com/keugenek/shark/tree/main/commands/shark-autotune Command: npx skills add https://github.com/keugenek/shark --skill shark-autotune
Manually configuring SHARK_LOOP_TIMEOUT and SHARK_MAX_LOOPS for Shark pattern AI coding agents relies on guesswork, leading to either frequent task timeouts from overly short settings or wasted idle time from overly long ones. This Skill eliminates that guesswork by analyzing actual historical timing data from your agent runs to deliver data-driven, tailored configuration recommendations.
Use the shark-autotune skill to analyze your Shark agent's timing history and get recommended timeout and loop configuration settings.
High-intent search queries and answers about installing and using this skill.
Analyze historical Shark agent timing data to calculate median and p95 turn times, timeout rates, and loops to completion metrics, eliminating configuration guesswork by providing tailored timeout and loop limit recommendations.
Your AI coding agent times out because generic SHARK_LOOP_TIMEOUT values are too short for your workload. Analyzing historical timing logs calculates actual p95 turn times and loops to completion, revealing the optimal timeout threshold.
The best way to configure loop limits is to process historical Shark agent timing logs to calculate loops to completion metrics, delivering evidence-based SHARK_MAX_LOOPS recommendations tailored to your specific task execution patterns.
Yes, historical timing analysis fixes excessive idle waiting by detecting wasted headroom from overly generous timeout configurations. It processes prior Shark run logs to recommend adjusted SHARK_LOOP_TIMEOUT values that eliminate unnecessary waiting.
Yes, you need prior task run logs. The tuning process analyzes historical Shark agent timing data from previous runs to calculate performance percentiles and timeout rates, generating data-driven configuration recommendations.
keugenekRemove Shark pattern state files while preserving timings.jsonl on confirmation.
davpatel605-beepDelegate software development tasks to an autonomous OpenCode CLI agent.
keugenekEnforce OS-level per-turn timeouts for AI coding agent loops.
mathiasbourgoinScan installed roster agent files and guide validated configuration edits.
