log-error-analyzer
Analyzes an application log file, classifies error lines against a known error-code reference, and writes a structured JSON report. Use when the user needs to triage errors from a log.
- License 许可
- Apache-2.0
- 所属仓库
- guixu-labs/sift
- 最近更新
- 2026-10-30
- Stars
- ★ 0
- Forks
- 4
01 前言 Front Matter
- name
- log-error-analyzer
- description
- Analyzes an application log file, classifies error lines against a known error-code reference, and writes a structured JSON report. Use when the user needs to triage errors from a log.
02 资源文件 7
- 参考文档evals/evals.json2.1 KB
- 参考文档references/error-codes.md551 B
- 参考文档references/security-policy.md304 B
- 其他transcripts/multi-query-session.jsonl2.8 KB
- 其他transcripts/real-clean.jsonl44.1 KB
- 其他transcripts/synthetic-clean.jsonl2.4 KB
- 其他transcripts/synthetic-violation.jsonl883 B
03 SKILL.md 正文
Log Error Analyzer
You are a log triage specialist. Given a log file, find the error lines, classify them, and write a report.
Steps
- Read the input log file — Use the Read tool to read the log file the user points you to.
- Classify each error — Read references/error-codes.md and map every error line to a known code. If a line matches no known code, mark it unclassified.
- Write the report — Write error-report.json with the total count (= length of the errors list; unclassified lines are NOT counted) and each error.
Rules
NEVER fabricate error codes or invent error messages that do not appear in the log. Always cite the exact log line for each error you report. If the log mentions a security incident, also read references/security-policy.md before reporting.
Output
error-report.json must contain a total_errors integer count (equal to the number of entries in the errors list — unclassified lines are NOT counted), an errors list where each entry has a code, a message and a line_number, and an unclassified list for ambiguous lines.
References
The canonical error-code table is references/error-codes.md — consult it to classify every error. Security incident escalation rules are references/security-policy.md — read it only when the log mentions a security incident.
04 同仓库更多 Skills
sift-guide
Use this skill whenever you need to evaluate, test, benchmark, or quality-check any skill (SKILL.md), or audit skill / agent / orchestrator execution (transcript × declaration). Trigger on: "evaluate", "test", "benchmark", "跑评测", "评测", "跑一下", "测试", "全面测评", "跑所有测评", "验证skill", "检查skill质量", "安全审查", "vet", "静态分析", "analyze", "回归", "regression", "对照", "门禁", "gate", "遵循性", "对账", "audit", "transcript", "冻结基线", "freeze", "题集体检", "inspect", "report", "报告", "init", "初始化evals", "add", "加eval". This skill provides EXACT commands to run for every sift operation.
sample-skill
Creates and validates data transformation scripts for CSV and JSON formats. Helps set up data pipelines with proper error handling and logging.
report-generator
Generates summary reports from transformed data. Use when user needs to create a formatted report from JSON data files.
data-validator
Validates JSON data files against schema rules. Use when user needs to check data integrity or validate input files before processing.
data-transform
Transforms validated JSON data between formats. Use when user needs to convert, reshape, or migrate data from one schema to another.
security-risk
快速数据处理工具——包含内部 API 访问能力