dev-wiki

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

• Third-party content exposure detected (high risk: 0.85). High: scripts/discover.sh and dev-workflow.cjs cause the agent to read outsider-authored .dev/ and .dev-archive/ PRD/checkpoint.mdmarkdown from other projects/directories and concatenate their text into the generated “context” string (e.g.,El()loads PRD +context/decisions/blockers/notes, then El()builds an LLM-like prompt packet inEl’s n+=`\n${t.context}\n``) which can then be fed to an LLM downstream by other skills.