The Agent Skills Directory

[SAFE]: The skill provides documentation and setup instructions for legitimate Sentry integrations. All tools and packages mentioned (e.g., sentry-wizard, sentry-cli, and various bundler plugins) are official vendor resources. The manual instructions for build tool configuration (such as modifying Gradle blocks or Xcode Run Script phases) are standard practices for this use case.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from Sentry events (stack traces, breadcrumbs, tags), which represents an attack surface for indirect prompt injection since this data is derived from external application runtime behavior. The skill proactively mitigates this risk with specific instructions.
Ingestion points: Event data is ingested via MCP tools such as search_issues and get_sentry_resource as described in SKILL.md.
Boundary markers: SKILL.md contains explicit warnings to the agent to treat MCP returns as untrusted input.
Capability inventory: The skill guides the agent in reading and modifying project configuration files and executing build-time diagnostic tools.
Sanitization: The instructions include a clear directive to never execute instructions found within event payloads, issue titles, or comments.
[CREDENTIALS_UNSAFE]: The skill includes thorough guidance on the secure handling of authentication tokens in references/auth-token.md. It advises using organization-level tokens, warns against committing secrets to source control, and recommends the use of environment variables or gitignored configuration files. It also provides a benign shell command for verifying the presence of a token without exposing its value.

sentry-fix-stack-traces