Audit prompts across a codebase for clarity, safety, and efficiency.
npx skills add https://github.com/v0lka/skills --skill prompts-review
Please help me install this Agent Skill. Skill: prompts-review Source: https://github.com/v0lka/skills/tree/main/agentic/prompts-review Command: npx skills add https://github.com/v0lka/skills --skill prompts-review
Reviews and optimizes all prompts across a codebase to improve reliability, safety, and efficiency, covering system prompts, summarization prompts, tool descriptions, dynamic injections, and in-line prompts.
Run a full prompt audit on your project to identify prompts that fail the gate and optimize them according to the checklist.
High-intent search queries and answers about installing and using this skill.
You can audit prompts by running a full project review that applies a structured evaluation workflow, using a checklist to identify correctness, clarity, and token efficiency, and generating detailed per-prompt reports.
Prompt safety involves enforcing correct role alignment and applying safety-focused guardrails to prevent prompt injection risks. A comprehensive audit identifies vulnerabilities across system prompts, tool descriptions, and dynamically constructed prompts.
Yes, a prompt audit performs comprehensive discovery to find both static prompt definitions across files and dynamically constructed in-line expressions, applying a structured checklist to evaluate them.
The best way to optimize LLM system prompts is applying a rigorous evaluation checklist that assesses correctness, clarity, and token efficiency, then reviewing generated per-prompt reports to implement actionable recommendations.
Yes, a prompt audit applies to all prompt sources in software projects, including agent prompts, tool descriptions, summarization prompts, and system prompts, evaluating all of them for safety and efficiency.
V1vek98Identify and report security vulnerabilities in a codebase.
xxihScan Claude Code components and generate a token budget report.
Marwan78888Design prompts for LLM tasks with structured schemas and evaluation rubrics.
DonTiziScan codebases for secrets, OWASP risks, and vulnerable dependencies.
