The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests data from external Sentry reports (issue details, events, and logs) which may contain attacker-controlled content embedded in error messages or application breadcrumbs. This data enters the agent's context through tools like mcp__sentry__get_issue_details and mcp__sentry__search_issues.
[Ingestion points]: External data enters via the mcp__sentry__ tool suite as defined in SKILL.md.
[Boundary markers]: There are no explicit delimiters or instructions provided to the agent to distinguish untrusted Sentry data from system instructions.
[Capability inventory]: The agent has the capability to search events, analyze issues with Seer, and read organization metadata.
[Sanitization]: While the skill includes instructions to redact PII and limit raw stack trace output, it lacks input validation or sanitization to prevent the agent from obeying instructions embedded within the ingested Sentry data.
[SAFE]: The skill promotes secure practice by explicitly instructing the agent to avoid direct use of SENTRY_AUTH_TOKEN and local Python scripts in favor of managed MCP tools.
[SAFE]: No unauthorized network operations, file system access, or remote code execution patterns were identified.

sentry