A collection of agent skills for AI-assisted development workflows.
Auto-split mixed changes into logical commit batches with validated Conventional Commit messages. Automatically intercepts all git add, git commit, and git push operations to enforce a plan-first workflow with built-in safety gates.
Key features:
- Plan-first workflow prevents accidental mixed commits
- Built-in safety gates (sensitive data, conflict markers, protected branches, large files, etc.)
- Auto-execute by default, plan-first on request
- Cross-platform support: Codex, Claude Code, Kiro, Kimi, Qwen, Gemini, OpenAI
Parse SonarCloud project, issue, and security hotspot links, then fetch normalized read-only details for agent analysis and code-fix workflows.
Key features:
- Supports project, issue, and security hotspot links
- Batch processing for multiple links
- Agent-friendly JSON and markdown output
- Read-only — never modifies SonarCloud state
Orchestrate a complete local repository quality-gate closure loop: audit, fix, and verify against AGENTS.md, local CI, Skills, reviews, specs, and quality gates, then produce an evidence-backed report.
Key features:
- Read-only probe collects git state, CI workflows, configs, stack, tools, and local Skills
- Quality Gate Manifest with evidence-based completion gate
- Scope control: focus on current branch/PR, avoid legacy debt cleanup
- Mode escalation requires explicit authorization
- Context-aware workflow parser: ignores
env.run/with.run, handles block scalar comments - Secret redaction (URL credentials, tokens, Bearer/Basic auth, private keys) and credential safety
--base-reffor accurate diff scope aligned with PR base (supports branch, tag, SHA)- Nested
AGENTS.mddiscovery for per-module agent instructions - JSON output includes
schema_version,probe_version,generated_at - Markdown paths normalize
$HOMEto~ - Extended config detection: Python, JVM, C/C++, containers, task runners
- C / NGINX-adjacent repository checks
- Requires Python 3.10+
npx skills add cnkang/skills/<skill-name>