// Search past decisions, failures, and experiment logs for relevant context before starting a task. Use this before any significant implementation or experiment to avoid repeating mistakes.
Analyze sweep results after completion. Computes accuracy, CFR, signal statistics, and generates comparison tables. Use after a sweep completes to understand the data.
Compare results across models (Qwen vs Llama) at matching compression configurations. Generates side-by-side tables and identifies cross-model patterns.
After completing a significant task or experiment, extract lessons learned and update the project knowledge base. Captures what worked, what failed, and what to remember for next time.
Prepare and run a KV-cache compression sweep. Loads sweep configuration, validates prerequisites, and provides the exact commands needed. Use before starting any GPU experiment.
| name | advise |
| description | Search past decisions, failures, and experiment logs for relevant context before starting a task. Use this before any significant implementation or experiment to avoid repeating mistakes. |
| argument-hint | ["topic"] |
| allowed-tools | Read, Grep, Glob |
Before starting work on "$ARGUMENTS", search the project's knowledge base for relevant past experience.
Decision log — docs/decisions.jsonl
Search for decisions related to the topic. Each line is a JSON object with: date, title, context, options, chosen, rationale, references.
Known failures — docs/failures.md
Check if there are documented failure patterns related to the task.
Experiment reports — docs/experiments/*.md
Look for past experiments that produced relevant findings.
SOTA analysis — docs/analysis/*.md
Check if the SOTA review covers this topic.
v2 plan — docs/v2-plan.md
Check if this task is already planned and what the requirements are.
Related work — docs/related-work-positioning.md
Check how other systems handle this.
Present findings as:
If no relevant context is found, say so clearly — absence of past context is also useful information.