The Agent Skills Directory

[SAFE]: The skill implements a robust safety workflow for destructive actions (suppressing error data). It includes mandatory dry-runs, blast-radius assessments, and explicit human-in-the-loop confirmation before creating rules.
[SAFE]: All external references and tools target official PostHog infrastructure (posthog.com) and standard error-tracking domains (e.g., browser extension namespaces), which is consistent with the skill's purpose as an administrative tool for PostHog users.
[COMMAND_EXECUTION]: The skill uses posthog:execute-sql for dry-running filters. While this allows SQL execution, it is scoped to a read-only preview of error data matching the user's intent, and the instructions specifically guide the agent to use it for safely testing suppression logic before deployment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted error data (exception messages, stack traces). However, it includes safety instructions to verify these patterns manually and use narrow filters to prevent 'schema confusion' where a malicious error message might trigger an overly broad suppression rule.

suppressing-noisy-errors