Results for “pruning”

Automate multi-turn adversarial conversations against LLM agents using Microsoft PyRIT, including Crescendo and Tree-of-Attacks-with-Pruning (TAP) attack chains with scorer feedback loops.

Optimize context usage for the session by reducing bloat, improving token efficiency, and focusing on relevant areas. Use when context is growing large, responses are slow, or switching between unrelated tasks. Covers context analysis, compacting, targeted pruning, MCP-first strategies, and focused scoping by area.

Inspect, plan, and invoke safe cleanup for runtime-manager-owned state without binding the workflow to a specific host orchestrator or agent provider.

Critically audit agent skills and remove content that is outdated, disproven, model-specific, or based on poorly cited sources. Load when improve-skills runs its per-skill cycle, when the user asks to prune skills, remove outdated techniques, check if skills are still valid, verify citations in skills, audit skill sources, or update skills for a new model release. Also triggers on "are these skills still valid", "check for obsolete techniques", "verify skill citations", or "update skills for GPT-5/Claude 4/Gemini 2". Runs before split-skill and compress-skill — removing bad content first means the remaining content is worth preserving.

Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.

Build and debug Retrieval-Augmented Generation pipelines — chunking, embedding, retrieval, reranking

Design, tune, and evaluate production RAG pipelines with deterministic tools for chunking, pipeline design, and retrieval evaluation.

Visual Prompt Tuning

Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.

Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.

Routes NuRec/Neural Reconstruction requests to the correct upstream NVIDIA skill (datasets, conversion, training, rendering, object harvesting, frame cleanup).

Train, evaluate, export, prune, and run inference for PointPillars 3D object detection models from LiDAR point clouds using NVIDIA TAO.

Moves issues and external PRs through a state machine of triage roles, categorizing, verifying, grilling, and writing agent-ready briefs.

Design or revise a company's commercial policy governing discounts, approver thresholds, exception flows, and deal frameworks using data-backed matrices and linting tools.

Manages vector embeddings with Pinecone for semantic search, recommendation, and RAG pipelines.

Use this skill whenever a developer wants to turn an idea into a complete, production-ready, end-to-end system plan BEFORE writing any code. Trigger on 'plan this system', 'design the architecture for', 'help me blueprint', 'deep plan for X', 'break this idea into components', 'expand into action points', 'full implementation plan', or when the user pastes a project idea wanting architecture, components, pipelines, and file-level execution mapped out. Casual phrasing also triggers: 'help me think this through end-to-end', 'plan before coding'. Also covers living-workspace patterns: self-improving knowledge bases, reflection loops with auditor agents, four-agent teams, schema-as-code, wiki health scoring. **Resume triggers**: 'where did we leave off', 'continue this plan', 'resume my blueprint' - rehydrates state from disk via SESSION.md/NEXT.md/decisions/. Web research is mandatory every invocation.

Turn a rough task into a precise, ready-to-paste prompt for any AI platform (Claude, GPT, Gemini, and others). Trigger on "write me a prompt for", "I need an AI to", "help me prompt engineer", "build a system prompt", "make this prompt better", "optimize my prompt", or any prompt-building request.

Rank a topic's user-flow branches by proof priority (value × risk × frequency) right after user-flow-map, ordering the branches, gating variation breadth, and promoting or pruning flows so state-model and ux-variations grow the tree in proof order — writes only existing flow-tree ordering fields, no schema change.

Diagnose and fix macOS resource pressure for Claude Code workflows. Use when load avg is high, swap is saturated, CC feels slow, or before spinning up parallel agents/worktrees. Covers CPU top-talkers, swap pressure, zombie claude processes, Node heap tuning, Spotlight/background-agent pruning, and purge. Apple Silicon aware.

Reduce OpenClaw token usage and API costs through smart model routing, heartbeat optimization, budget tracking, and native 2026.2.15 features (session pruning, bootstrap size limits, cache TTL alignment). Use when token costs are high, API rate limits are being hit, or hosting multiple agents at scale. The 4 executable scripts (context_optimizer, model_router, heartbeat_optimizer, token_tracker) are local-only — no network requests, no subprocess calls, no system modifications. Reference files (PROVIDERS.md, config-patches.json) document optional multi-provider strategies that require external API keys and network access if you choose to use them. See SECURITY.md for full breakdown.

Creates, updates, or optimizes an AGENTS.md file for a repository with minimal, high-signal instructions covering non-discoverable coding conventions, tooling quirks, workflow preferences, and project-specific rules that agents cannot infer from reading the codebase. Use when setting up agent instructions or Claude configuration for a new repository, when an existing AGENTS.md is too long, generic, or stale, when agents repeatedly make avoidable mistakes, or when repository workflows have changed and the agent configuration needs pruning. Applies a discoverability filter—omitting anything Claude can learn from README, code, config, or directory structure—and a quality gate to verify each line remains accurate and operationally significant.