The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The script scripts/check-sentry.sh explicitly suggests a piped remote execution pattern for installation: curl -sL https://sentry.io/get-cli/ | bash. This is a classic RCE vector that bypasses package manager safety checks and executes unverified remote content directly in the shell.
[CREDENTIALS_UNSAFE] (HIGH): The skill is designed to read and export sensitive authentication data from ~/.sentryclirc and environment variables. The lib.sh helper and scripts/check-sentry.sh logic retrieve SENTRY_AUTH_TOKEN, exposing raw API credentials to the agent's processing environment.
[COMMAND_EXECUTION] (MEDIUM): scripts/sentry-api.sh provides a wrapper for making authenticated HTTP requests. While it attempts to block the DELETE method, it allows arbitrary GET, POST, and PUT requests to the Sentry API using placeholders. This creates a surface where a malicious prompt could trick the agent into performing unauthorized state-changing operations on the user's Sentry projects.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill performs multiple unpinned network operations using curl to https://sentry.io. While targeting a reputable service, the lack of integrity checks on the downloaded content (especially in the suggested installation command) poses a supply-chain risk.
[DATA_EXFILTRATION] (LOW): By design, the skill fetches issues and event data from Sentry. While this is the intended purpose, it involves pulling potentially sensitive application error data (including stack traces and user context) into the agent's context.

cli-sentry