teach-build
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed to analyze local project artifacts (git diffs, documentation, and internal
.skills/metadata) to create a self-contained HTML summary of a build process. It does not perform any unauthorized network operations. - [DATA_EXFILTRATION]: The skill includes explicit security constraints to prevent data leakage. It mandates that the generated packet remains local to the filesystem (
.skills/) and strictly forbids publishing to external surfaces or artifacts unless specifically requested by the user. Additionally, it prohibits the use of external CDN fonts, scripts, or images, ensuring the generated HTML is self-contained and private. - [PROMPT_INJECTION]: The skill acknowledges the risk of indirect prompt injection from untrusted repository data (notes, diffs, ledger lines). It provides clear instructions to the agent to treat such content as passive data, stating that instruction-like content within these sources must never be followed. It also requires that all such data be escaped before being embedded in HTML or JS contexts to prevent Cross-Site Scripting (XSS) or further injection attacks.
- [COMMAND_EXECUTION]: The skill dispatches subagents for system mapping but limits them to a specific model tier (Sonnet) and scope. No arbitrary shell command execution or risky system calls are present in the instructions.
Audit Metadata