The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes multiple shell commands including gh, vercel, and curl to gather production data. It also attempts to execute locally stored scripts from ~/.claude/skills/triage/scripts/.
[DATA_EXFILTRATION]: Reads configuration from .env.local to determine the application's health endpoint. While used for diagnostics, this demonstrates access to environment configuration files which may contain sensitive secrets.
[PROMPT_INJECTION]: The skill processes untrusted external data from Sentry errors and Vercel logs which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Sentry issue reports, Vercel log output, GitHub Action run details.
Boundary markers: Absent.
Capability inventory: Shell command execution (bash, gh, vercel, curl), file system reads (grep).
Sanitization: Absent.

check-production