systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOW
Full Analysis
- DATA_EXPOSURE (INFO): In Phase 1.4, the skill provides shell examples for gathering evidence in multi-component systems, including
security list-keychainsandsecurity find-identity. While these access system security metadata, they are presented as standard troubleshooting steps for build and signing failures. The skill proactively suggests a safe pattern for checking environment secrets (${IDENTITY:+SET}) that confirms a variable is set without printing its value. - INDIRECT_PROMPT_INJECTION (LOW):
- Ingestion points: Phase 1 involves reading error messages, stack traces, and component logs from potentially untrusted external systems.
- Boundary markers: Absent; the skill does not explicitly define delimiters for untrusted data.
- Capability inventory: Execution of diagnostic shell commands (
grep,security,codesign) and decision-making regarding code fixes. - Sanitization: Absent; however, the skill is a reasoning framework for a human-agent loop rather than an autonomous execution pipeline, significantly mitigating the risk of obedience to adversarial content embedded in error logs.
Audit Metadata