repo-report-builder
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to ingest and analyze external repository source code.
- Ingestion points: Repository cloning and inventory workflows (SKILL.md).
- Boundary markers: None specified to delimit ingested code content from agent instructions.
- Capability inventory: Cloning/fetching repositories, spawning sub-agents for analysis, executing tests/builds/audits, and utilizing
browser-use:browserfor report verification. - Sanitization: The skill includes a 'Redaction' policy for output generation but does not specify sanitization or escaping techniques for the input code analysis phase.
- [COMMAND_EXECUTION]: The workflow allows for the execution of repository-specific tests, builds, and audits. This is a primary function of the skill's auditing capability and is explicitly gated by a discovery question (Question 7) requiring user permission before any dynamic execution occurs.
- [EXTERNAL_DOWNLOADS]: The skill mentions using a Mermaid CDN to render diagrams in the generated HTML reports. This is a standard practice using well-known public infrastructure for visualization.
Audit Metadata