result-diagnosis
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it instructs the agent to ingest and analyze potentially untrusted data sources while maintaining access to high-capability tools like Bash and file system modification.
- Ingestion points: The skill reads logs, configuration files, metric files, and checkpoints (
references/evidence-audit.md), and fetches external content via web search and web fetch (SKILL.md). - Boundary markers: Absent. There are no instructions to use delimiters or specific warnings to ignore embedded instructions within processed data artifacts.
- Capability inventory: The skill has access to
Bash,Write,Edit,Read, andGlobtools (SKILL.md), allowing it to execute shell commands and modify the project environment based on its findings. - Sanitization: Absent. There is no mention of validating, filtering, or escaping content from logs or external documents before processing.
Audit Metadata