The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external data like error messages, stack traces, and system logs, which constitutes an attack surface for indirect prompt injection. This is a functional requirement for a debugging skill and no exploitation patterns were observed.\n
Ingestion points: Phase 1 (Root Cause Investigation) involves reading error messages, stack traces, and component logs from the execution environment.\n
Boundary markers: The skill does not mandate the use of delimiters when reading external content.\n
Capability inventory: The skill utilizes command execution (npm, security, bash) and file system access.\n
Sanitization: The instructions do not include specific sanitization steps for the ingested logs.\n- [COMMAND_EXECUTION]: The skill includes a utility script and examples that perform shell operations to diagnose system state.\n
Evidence: The find-polluter.sh script executes npm test on local files. SKILL.md provides examples of using macOS security and codesign utilities for debugging build and signing workflows.

systematic-debugging