dify-docs-feature-research
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands to manage repositories and fetch data.
- Evidence: Use of
git fetch,git checkout,git pull,grep, and the GitHub CLI (gh issue list,gh search issues,gh api) to interact with the local environment and remote repositories. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data that could influence the agent's behavior or output.
- Ingestion points:
SKILL.mdidentifies GitHub issues and discussions fromlanggenius/difyandlanggenius/graphonas primary data sources for Phase 2. - Boundary markers: No explicit boundary markers or instructions to ignore embedded instructions within the fetched GitHub content are provided.
- Capability inventory: The agent has the ability to execute shell commands (
git,gh,grep) and read local file contents. - Sanitization: There are no instructions to sanitize or escape the user-provided feature names or the community feedback before they are used in shell commands or synthesized into the research report. This creates a surface where a malicious GitHub issue or a specially crafted feature name could attempt to manipulate the agent's command execution or final summary.
Audit Metadata