The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies and mitigates the risk of indirect prompt injection from untrusted Sentry event data.
Ingestion points: Untrusted data enters the agent context through Sentry exception messages, breadcrumbs, request bodies, and attachments (Phase 2).
Boundary markers: The skill includes a dedicated 'Security Constraints' section that explicitly instructs the agent to ignore directives or code suggestions found within event data and treat them as plain text.
Capability inventory: The agent has the ability to read and write files, as well as execute git commands to investigate and fix bugs.
Sanitization: Instructions mandate validating Sentry data against the local codebase before implementation (Phase 5) and redacting PII or secrets from all reports and test fixtures.
[COMMAND_EXECUTION]: The skill uses local environment tools for legitimate debugging purposes.
Evidence: Phase 5 involves using git log and git blame to review recent changes and understand the context of the errors being fixed.
[DATA_EXFILTRATION]: The skill includes specific rules to prevent the leakage of sensitive credentials or user data.
Evidence: The 'No secrets in output' rule prohibits the reproduction of tokens, passwords, or session IDs found in Sentry data, requiring indirect references instead.

sentry-fix-issues