1.8 KiB
TASK encryption-threat-model: per-note and secret-folder design only
Runs through OpenCode (specs/agents/opencode-runbook.md is attached automatically). Task 1 of specs/plugins-parity/note-encryption/tasks.md.
OWNS
specs/plugins-parity/note-encryption/threat-model.md(new, FIRST; only writable file)
FILES TO OPEN
specs/plugins-parity/note-encryption/spec.mdD1crates/engine/src/vault/session.rscrates/vault/src/lib.rs
GOAL / API
Write specs/plugins-parity/note-encryption/threat-model.md in SECOND tool call with initial acceptance checklist; first call reads at most three listed inputs. Document-only task: no code stub or running crypto.
Produce asset/adversary/trust-boundary matrix for per-note .md.enc and proposed opaque secret-folder domain. Distinguish offline locked-vault theft, sync storage, unlocked/compromised host, rollback/deletion, path/size/timestamp leakage, shared attachments and historical plaintext. Separate local at-rest lock from sync E2EE. Trace protected/excluded/blocked states to source boundary. Q1/Q4 pending owner decisions remain explicit; no implementation authorized.
TESTS
Tabletop cases: stolen locked vault, replay old ciphertext, swap encrypted objects, hostile same-user process while unlocked, outside plaintext attachment, known filename and old backup. Every case maps to protection/exclusion/blocker with reason; no false secure-erasure/zero-knowledge claim.
VERIFY
Use rg --files to confirm inspected source paths; read resulting document against spec D1 and acceptance 1/4. Confirm only owned document changed. No cargo or crypto execution for design task.
DO NOT
Code/Cargo/git changes, dependencies, crypto implementation, secure-erasure promise, choose algorithm on owner's behalf, mark owner approval granted.