The Agent Skills Directory

[DATA_EXFILTRATION]: The skill's debugging instructions (SKILL.md) suggest adding instrumentation that prints environment variables and keychain status (e.g., env | grep IDENTITY, security list-keychains). While intended for local troubleshooting of signing and build workflows, these actions can expose sensitive configuration data in the agent's output or logs.
[COMMAND_EXECUTION]: The file find-polluter.sh is a utility script that automates running pnpm test on discovered test files. This is standard developer tool behavior but represents local command execution based on file system contents.
[PROMPT_INJECTION]: The skill's primary function is to analyze untrusted data sources like error messages, stack traces, and component logs (SKILL.md Phase 1). This creates a surface for indirect prompt injection, where an attacker could influence the agent's behavior by embedding instructions in the logs or error output being investigated.
Ingestion points: Error messages, stack traces, and diagnostic logs (SKILL.md Phase 1).
Boundary markers: Absent.
Capability inventory: Subprocess calls for bash commands, pnpm test, and system utilities like codesign (find-polluter.sh, SKILL.md).
Sanitization: Absent.

systematic-debugging