用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/fabioc-aloha/Alex_Skill_Mall --skill symfix命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | symfix |
| description | Classify and remediate CodeQL alerts with the SymFix workflow using staged changelog updates |
| lastReviewed | 2026-04-30T00:00:00.000Z |
Scope: Classify and remediate the requested CodeQL alert set with SymFix
Use SymFix to classify and remediate the requested alert set with the minimum safe repository diff.
SymFix runtimes may expose tool names that differ slightly from the labels used in this guide.
Before Step 1, confirm all of the following:
symfix_staging_directory in the user profile directory and use it as the staging directory.SymFixGetCodeQLAlerts to get the list of CodeQL alerts.For each alert in the list, follow these sub-steps in order:
SymFixUpdateAlert to set up the code graph for the alert and update the alert line numbers.SymFixUpdateAlert. For the first alert, use an empty list. For each subsequent alert, reuse the accumulated changelog ranges returned by earlier applied fixes.SymFixGenerateAlertClassificationPrompt to get the alert classification prompt. Follow the prompt instructions. Use the SymFix code navigation tools GetContainingDefinition, GetDefinition, GetMethodCallers, GetMethodCallees, GetReferences, and GetImports as needed.SymFixValidateAlertClassificationResponse to validate your response to the alert classification prompt.SymFixGenerateAlertFixPrompt to get the fix generation prompt. Follow the prompt instructions.SymFixValidateAndApplySearchReplaceChangelog to validate and apply the SearchReplace changelog that is your response to the fix prompt. This tool takes the earlier changelog ranges list as input, updates it with the current fix, and returns the updated list.For each alert, end the SymFix phase in exactly one of these states before moving on:
FIXEDFALSE_POSITIVEFAILEDRecord the state for every alert so the final output can report exact per-alert results.