chromadb-integration-skills
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides high-quality documentation and code examples for database integration, emphasizing defensive programming. It includes dedicated modules for error resilience, testing, and security best practices such as PII redaction and secret sanitization. All external code patterns are standard library usage or references to well-known services.- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for ingesting untrusted data from external sources (e.g., research papers, GitHub issues, documentation) into a vector database for semantic search. This establishes a surface for indirect prompt injection if retrieved content is used in subsequent LLM prompts.
- Ingestion points: The skill describes fetching data from APIs and files in SKILL.md and chromadb-lifecycle-management.md.
- Boundary markers: While interpolation examples use clear naming, they do not explicitly enforce structural delimiters (like XML tags) to isolate retrieved content from instructions within the provided snippets.
- Capability inventory: The skill utilizes the mcp__chroma__* tool suite for database operations.
- Sanitization: The skill specifically addresses this risk through the chromadb-security-patterns.md module, which provides robust logic for redacting PII, secrets, and system paths before storage, effectively mitigating the threat.
Audit Metadata