lsp-architecture
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase metadata to generate architecture overviews, creating a surface where malicious symbol names or file paths could influence agent behavior.
- Ingestion points: Codebase symbols, directory structures, and file names are ingested in Steps 1 through 4 using the
mcp__lsp__find_symbolandmcp__lsp__list_symbolstools. - Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" warnings for the data retrieved from the workspace.
- Capability inventory: The skill's capabilities are limited to read-only LSP operations; no tools for file modification or network communication are permitted in the frontmatter configuration.
- Sanitization: No explicit sanitization or validation is performed on the symbol names or file paths retrieved from the local environment before they are included in the final report.
Audit Metadata