skills.rest
catlog22
catlog22COMMUNITY
Explore all skills by this author →
GitHub

issue-discover

Creates and discovers code issues via multi-perspective analysis and prompt-driven exploration.

2.1k|165|Updated Sep 7, 2025
Search Tags:#static analysis#code discovery#gitHub#Security audit#Multi-Agent#Bug detection#Issue tracking
One-click install
npx skills add https://github.com/catlog22/Claude-Code-Workflow --skill issue-discover
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: issue-discover
Source: https://github.com/catlog22/Claude-Code-Workflow/tree/main/.codex/skills/issue-discover
Command: npx skills add https://github.com/catlog22/Claude-Code-Workflow --skill issue-discover

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Manually auditing code for bugs, security flaws, and test gaps is slow and inconsistent. This Skill unifies issue creation and discovery into one workflow: create issues from GitHub URLs or text, run multi-perspective code analysis, or explore the codebase iteratively from a natural-language prompt.

Core Features & Use Cases

  • Issue Creation: Create structured issues from GitHub URLs, issue numbers, or plain text with clarity-based clarification and optional GitHub publishing.
  • Multi-Perspective Discovery: Analyze target files across 8 perspectives (bug, security, test, quality, performance, UX, maintainability, best-practices) using parallel subagents, with optional Exa external research.
  • Prompt-Driven Exploration: Describe what to find in natural language; Gemini plans an exploration strategy and iterative agents investigate with ACE semantic search, including cross-dimension comparison (e.g., frontend vs backend API contracts).
  • Quick Plan & Execute: Convert high-confidence findings directly into executable task files and fix them inline, skipping full issue registration.
  • Use Case: Run a security-focused discovery on src/payment/**, review the prioritized findings, then either export them as tracked issues or immediately fix the critical ones.

Quick Start

Ask the assistant to run issue-discover on a path like src/auth/** to discover potential bugs and security issues, or provide a GitHub issue URL to create a tracked issue.

Frequently Asked Questions about issue-discover

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I discover bugs and security issues in a code directory?▼

Run issue-discover with the discover action and a file glob pattern, such as src/auth/**. It spawns parallel analysis agents across perspectives like bug, security, and test, then aggregates prioritized findings into issue candidates.

How do I create a GitHub issue from a URL or text description?▼

Pass a GitHub issue URL, an issue number like #42, or a plain text description to issue-discover. It fetches or parses the content, optionally asks clarifying questions for vague input, and creates a structured issue via the ccw issue CLI.

What is the difference between discover and discover-by-prompt modes?▼

Discover mode uses fixed pre-defined perspectives (bug, security, test, etc.) executed in parallel. Discover-by-prompt accepts a natural-language goal, uses Gemini to plan an exploration strategy, and runs iterative agent exploration with ACE semantic search.

Can issue-discover fix the problems it finds automatically?▼

Yes, the Quick Plan & Execute phase converts high-confidence findings (confidence >= 0.7, critical or high priority) into task files and executes fixes inline. Lower-confidence findings can be exported to the issue tracker for full planning instead.

Does issue discovery support external security research?▼

Yes, Exa-based external research is automatically enabled for the security and best-practices perspectives, and can be forced on for all perspectives with the --external flag. Results are written to external-research.json in the discovery output directory.

What happens when discovery agents time out?▼

A four-step cascade handles timeouts: a status probe requests progress, then a force-finalize interrupt asks agents to output current findings, and finally the workflow continues with completed results. All agents are explicitly closed to prevent resource leaks.

Related Skills

Explore More Software Engineering Skills→
Atypical-ConsultingAtypical-Consulting
2

ghs-issue-analyze

Analyze GitHub issues by inspecting the codebase and posting structured analysis comments.

Official
Advanced
AsiaOstrichAsiaOstrich
6

orchestrate

Execute DAG-based task plans with parallel layers and execution_report.json output.

Community
Advanced
interactive-incinteractive-inc
1

backlog

Record, update, and plan tasks within a project backlog.

Official
Intermediate
DisentinelDisentinel
36

grafema-codebase-analysis

Analyzes codebases by building and querying a graph database from static analysis.

Community
Advanced