Score improvement proposals and emit retro artifacts with stop and promotion criteria.
npx skills add https://github.com/Mr-Fang-VLSI/EDAgent --skill eda-retro
Please help me install this Agent Skill. Skill: eda-retro Source: https://github.com/Mr-Fang-VLSI/EDAgent/tree/main/skills/eda-retro Command: npx skills add https://github.com/Mr-Fang-VLSI/EDAgent --skill eda-retro
This skill provides structured post-experiment retrospectives to identify root causes, categorize failure/success mechanisms, and decide if recursion to a new eda-loop iteration is warranted.
After each experiment batch, run the eda-retro skill to generate a retro summary and decide whether to trigger the next eda-loop.
High-intent search queries and answers about installing and using this skill.
EDA retrospectives are automated by consolidating run summaries, monitors, and manifests to identify root causes and categorize failure or success mechanisms. This structured review processes raw experiment batch artifacts to generate a clear narrative of what happened and why.
The best way to gate EDA recursion is by scoring improvement proposals against evidence-driven policies. The retrospective evaluates experiment batch data to enforce risk-aware decisions, emitting explicit stop and promotion criteria to determine if a new iteration is warranted.
Recursive actions are gated by applying evidence-driven recursion policies to post-experiment data. The process scores improvement proposals based on consolidated run summaries and manifests, ensuring subsequent iterations remain data-driven and explicitly defining stop criteria.
A post-experiment retro artifact should include hypothesis validation results, explicit stop and promotion criteria, and concrete next-step actions. It records the root cause analysis and categorizes mechanisms to guide whether a new EDA loop iteration should be triggered.
A structured retrospective is needed after completing an EDA experiment batch to consolidate raw artifacts. It is required when you must evaluate hypothesis validation, categorize failure mechanisms, and decide if evidence supports triggering a new recursion iteration.
okgoogle13Run sequential diagnostic scripts and report project health summaries.
Mr-Fang-VLSIMap legacy eda-chief calls to current agent policies and eda-loop execution.
Mr-Fang-VLSIAnalyze EDA route and timing evidence to produce a go/no-go decision.
niznik-devOrchestrate SLURM fine-tuning jobs with torchtune and evaluate with inspect-ai.
