The Agent Skills Directory

[SAFE]: The skill documentation and scripts are focused on improving debugging efficiency and software quality through systematic investigation and verification protocols.- [COMMAND_EXECUTION]: The script scripts/find-polluter.sh executes shell commands like npm test and ls. It uses proper shell quoting for variables, which prevents command injection from malicious file names.- [DATA_EXFILTRATION]: No network access or data exfiltration patterns were observed in the instructions or scripts.- [INDIRECT_PROMPT_INJECTION]: The skill involves the agent processing external data such as error messages and logs, which is an inherent surface for indirect prompt injection in debugging tasks. However, no malicious instructions or exploitation paths were identified.
Ingestion points: Error messages, stack traces, and log files analyzed in Phase 1 (references/systematic-debugging.md).
Boundary markers: Absent; the agent reads error data directly as provided by the system.
Capability inventory: Shell execution via scripts/find-polluter.sh (executing npm test).
Sanitization: Absent; the skill relies on the agent's internal safety filters.

Debugging