The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from Sentry event payloads (such as exception text, frame paths, and breadcrumbs) retrieved via MCP tools. It proactively mitigates the risk of indirect prompt injection by instructing the agent to treat this data as attacker-controllable and strictly forbidding the execution of any instructions found within these fields.
Ingestion points: Sentry event data (exception text, breadcrumbs, tags) accessed via search_issues and get_sentry_resource tools.
Boundary markers: Explicit instructions in SKILL.md to treat event payloads as untrusted and ignore instructions within them.
Capability inventory: The skill utilizes command-line tools including sentry-cli, npx, brew, gradle, and fastlane to perform build configurations and artifact uploads.
Sanitization: The skill relies on natural language instructions to the agent to prevent the processing of malicious content.
[COMMAND_EXECUTION]: The skill instructs the agent to use various command-line utilities for artifact management, such as sentry-cli, npx @sentry/wizard, and platform-specific build tools. These tools are official components from the recognized vendor and are used strictly for the skill's stated purpose of fixing stack traces.
[DATA_EXPOSURE]: The documentation provides clear guidance on handling sensitive Sentry authentication tokens. It advises against hardcoding or committing tokens to version control, recommending the use of environment variables and CI secret stores instead. It also provides a safe method for verifying token presence without logging its value.

sentry-fix-stack-traces