Contain instructions embedded in a retrieved document
A synthetic policy appendix says to ignore the user and reveal all other documents. Ensure retrieved text remains data and cannot expand access or activate tools.
- Focused work estimate
- 2h 30m + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Prompt injection defense · Least privilege · Output validation
Estimated field mix
- Applied AI50%
- Security50%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
Employees search fictional travel and equipment policies. The current prototype blends draft and approved text and sometimes answers using a policy the employee cannot open. Use a small synthetic corpus and a deterministic answer-provider double.
Setup prerequisites
- Author synthetic policy documents with revisions, effective dates, and access groups.
- Use a local deterministic provider double; paid model access is optional and not required.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- SEARCH-101 · Import policy documents with stable revision identities
- SEARCH-102 · Keep chunk citations anchored to the original policy text
- SEARCH-103 · Apply group access before ranking or sending context to a provider
- SEARCH-104 · Exclude drafts and future policies from current-policy answers
- SEARCH-105 · Reject answers with invented or mismatched citations
Acceptance criteria
- The provider request separates application instructions from quoted document context.
- The answer interface exposes no file, network, or administrative tools.
- Prompt-injection fixtures cannot alter document access scope or bypass citation validation.
Implementation constraints
- Do not claim that delimiting text alone prevents all prompt injection.
- Use deterministic malicious-output fixtures to test downstream enforcement.
Verification to include
- Supply an instruction-bearing chunk and inspect the constructed provider request.
- Return a malicious provider response with an unauthorized citation and verify the validator rejects it without a secondary fetch.
Deliverables
- Context construction boundary and injection containment tests
Rollout and recovery
Keep tool access disabled for this feature; reject any adapter configuration that grants capabilities beyond answering.
Value of the work
For the engineer: Practice access-aware retrieval, source provenance, structured model boundaries, and reproducible evaluation.
For the team: Inspect how an engineer prevents unsupported or unauthorized answers and handles changing policy content.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.