The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from the Sentry MCP (such as event payloads, issue titles, and comments) that originates from external application events. The instructions explicitly direct the agent to treat all MCP data as untrusted input and to never execute instructions found within these payloads, effectively mitigating the risk of indirect prompt injection from malicious event data.\n- [COMMAND_EXECUTION]: The skill utilizes official vendor CLI tools like sentry-cli and sentry-wizard to perform configuration and verification tasks. These commands are executed within the scope of the developer's project and are necessary for the skill's stated purpose of instrumentation.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing Sentry SDKs and build tools from official package registries (NPM, PyPI, etc.) and vendor repositories. These downloads are within the expected functional scope of a monitoring integration and target trusted, well-known sources.\n- [CREDENTIALS_SAFE]: The skill includes explicit guidance on the safe handling of authentication tokens, recommending the use of environment variables and gitignored configuration files to prevent the accidental exposure of secrets in version control.

sentry-instrument