Sentry Skills
github.com/getsentry/skills- Category
- AI Agents
- Rank
- No. 612Tools index
- Pricing
- Open Source
- Platform
- cli
- Type
- TOOL
- Builder
- @getsentry
- GitHub
- 985 stars
- Date
About
Agent Skills collection used by the Sentry team for development workflows. Provides pre-built skills for code review, commit management, security audits, and other development tasks that can be used with Claude Code, Cursor, and other AI agents.
What it does
Sentry Skills turns team practices into instructions that compatible coding agents can discover and apply. Its catalog spans pull request iteration, code quality, security, documentation, typing, onboarding, writing, and internal operational work. Several workflows include Python helpers that gather GitHub state and return structured JSON.
Why it's ranked here
The collection is unusually concrete: it pairs broad engineering guidance with operational helpers for pull request checks, review feedback, thread replies, and skill security scanning. The main caveat is deliberate specialization. Many entries encode Sentry conventions, systems, branding, or internal processes rather than general-purpose defaults.
What's good
The pull request tooling distinguishes actionable checks from human approval gates, extracts failure context, categorizes feedback, and reports structured outcomes. The security scanner checks frontmatter, prompt injection patterns, obfuscation, secrets, dangerous scripts, and unrestricted permissions. Contributor guidance also separates runtime instructions from maintenance contracts and reference material.
Tradeoffs
Some skills assume GitHub CLI access, specific organization conventions, external documentation, or Sentry infrastructure knowledge. The repository explicitly directs project setup and production debugging elsewhere. Skill testing is primarily described as local installation, invocation, behavior checks, and self-review, while formal review remains optional.
How to use it well
It best suits teams that want agent behavior to reflect written engineering policy across reviews, commits, pull requests, security checks, and documentation. Adopt individual skills, then customize organization-specific assumptions and permissions. Use another integration for installing Sentry in applications or debugging production incidents.
Technical notes+
README.md defines the Agent Skills packaging model, Claude Code marketplace installation, skills.sh installation, canonical skills tree, and two subagents. AGENTS.md requires SKILL.md plus SPEC.md for new or materially changed skills and documents registration rules. skills/skill-scanner/scripts/scan_skill.py is a Python 3.9 static analyzer using PyYAML and emitting JSON findings. The helpers under skills/iterate-pr/scripts/ use the gh CLI, REST and GraphQL calls, subprocesses, polling, and JSON output to inspect checks and feedback or reply to threads. skills/gh-review-requests/scripts/fetch_review_requests.py requires Python 3.12. .claude/settings.json contains an explicit command, fetch-domain, and skill allowlist. warden.toml runs skill-scanner for pull requests and locally, failing on high severity and reporting medium severity.
Observed
- License
- Apache License 2.0
- Primary implementation
- Markdown skill instructions with Python helper scripts
- Install surface
- Claude Code marketplace plugin or skills.sh package via npx
- Agent platforms
- Claude Code, Cursor, Cline, GitHub Copilot, and other Agent Skills-compatible agents
- Interfaces
- Agent skills and subagents, plus helper command-line scripts that emit JSON
- Runtime requirements
- Python helpers require Python 3.9 or 3.12; the scanner declares PyYAML
- Repository structure
- Canonical skills live at repository root, with separate agent definitions and maintenance contracts
Read from README.md, skills/skill-scanner/scripts/scan_skill.py, skills/iterate-pr/scripts/fetch_pr_checks.py, skills/iterate-pr/scripts/reply_to_thread.py, skills/iterate-pr/scripts/fetch_pr_feedback.py, skills/iterate-pr/scripts/monitor_pr_checks.py, skills/gh-review-requests/scripts/fetch_review_requests.py, LICENSE, AGENTS.md, CLAUDE.md, warden.toml, CONTRIBUTING.md, agents/senpai.md, .claude/settings.json.
What it can do
Review code for quality and best practices
Code files or pull requests → Code review feedback and suggestions
Perform security audits on code
Source code files → Security vulnerability reports and recommendations
Manage Git commits
Code changes and commit messages → Formatted commits and Git operations
Execute Git operations
Git commands and repository state → Git status updates and branch management
Analyze development workflow processes
Development workflow data → Process optimization recommendations
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.