Results for “safety-alignment”

Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.

Alignment

Enforces bulk-safety rules, sharing model requirements, CRUD/FLS security, SOQL injection prevention, PNB test coverage, and modern Apex idioms for Salesforce development.

Creates safety alert decks with hazard stripes, tiered cards, and policy code blocks for incident reviews and risk communication.

Use this skill for transporting steel coils. Covers § 393.120 specific rules, chain requirements, dunnage placement, and inspection procedures.

Create a consistent spacing system based on a base unit with contextual application rules for harmonious interfaces.

Reduce SIEM alert fatigue by systematically tuning detection rules in Splunk and Elastic, using statistical baselines, whitelists, and precision/recall metrics.

Develop and implement OT-specific incident response playbooks aligned with SANS PICERL framework, IEC 62443, and NIST SP 800-82 that address unique ICS challenges including safety-critical systems, limited downtime tolerance, and coordination between IT SOC, OT engineering, and plant operations teams.

当需要设定人生方向或进行重大行为改变时

Detect APT lateral movement by chaining Windows authentication events, process execution telemetry, and network connection logs across hosts using Splunk SPL and Sigma rule format.

Use when performing multiple sequence alignment of FASTA inputs, generating alignment ensembles, or calculating alignment confidence metrics.

当寻求个人行动或项目在更宏大系统中的意义，或需要将日常工作与长远价值连接时

Validate whether an OSS tool is worth building through staged namespace, existing-solution, and feasibility review

当设定新目标或检视现有目标时，确保目标能通过高效的行为系统自动达成

Use this skill for transporting pipes (water, gas, oil, structural). Covers § 393.122 specific rules and roll-prevention.

Skeleton loading placeholders — gray-shape stand-ins that mirror final content dimensions, use subtle shimmer/pulse animation, and prevent layout shift. Use when replacing generic spinners with content-shaped placeholders, reducing perceived latency on slow routes, or preventing CLS on image-heavy cards.

Use when implementing unified position sizing, Kelly criterion, or combining GARCH+Capital+Portfolio risk systems

Quick feasibility and novelty check — go/no-go decision for this week's build

Design loading, skeleton, and progressive content reveal patterns that maintain user confidence and perceived performance.

Improve layout, spacing, and visual rhythm. Fixes monotonous grids, inconsistent spacing, and weak visual hierarchy. Use when the user mentions layout feeling off, spacing issues, visual hierarchy, crowded UI, alignment problems, or wanting better composition.

Train AI models to be harmless through self-critique and AI feedback using a set of constitutional principles, without requiring human labels for harmful outputs.