gds-generate-project-context

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The workflow explicitly tells the agent to "load the referenced contents as facts" from file: entries and to carry those facts for the rest of the run — which can cause the LLM to ingest secrets from project files and potentially include them verbatim in generated outputs.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). Yes: customize.toml sets workflow.persistent_facts to file:{project-root}/**/project-context.md, and Step 3 loads those referenced file contents as verbatim facts into the agent’s LLM context at activation time; those project-context.md files are outsider-authored if they come from other people/commits in the project.