Tianyi Cui
cfceb8452b
The parent implementation introduced sandboxMode and approvalPolicy as generic SessionHeader fields, then propagated those fields through both persistence backends, session-query indexes, collision checks, policy-specific seed-boundary folds, catalogs, and a broad test matrix. That storage plane is unnecessary: Session already accepts a validated constructor seed, and persistence captures that seed when the session is announced before committing its first batch. Capture each parent override synchronously at delegation, append source-tagged sandbox/mode and approval/policy records after the optional fork prefix, and create the child with that combined seed. Keeping header.seedLength at the original fork-prefix length preserves lineage while ordinary last-event-wins folds make the inherited records outrank stale parent history and remain subordinate to later child switches. Unswitched parents still stamp nothing, so children continue to follow deployment defaults. Remove the generic header fields and every persistence/query/schema branch built around them. Collapse the inheritance suite from ten leaking scenarios to four owned-context cases covering real filesystem confinement, stale fork precedence, delegation-time capture, and the no-override path. The assembled headless snapshot now asserts the persisted inheritance event directly. This keeps the security behavior while restoring policy ownership to the existing event log and deleting the speculative durability machinery that the original tests did not exercise.
@deepseek-ai/dsh-user-approval
English | 中文
Channel-neutral one-shot approval seam. ctx.approval.request(req) returns allowed-once, rejected, cancelled, or unavailable; missing or failing answerers fail closed, and a grant applies only to the requested action. Exact event signatures live in the generated Cordis catalog.
Each request must belong to an open agent turn. The service appends a paired approval/asked and approval/decided audit record, while the model sees only the resulting logged tool outcome. An aborted request resolves cancelled; an audit append that fails before commit rejects rather than returning an unlogged decision.
Answerers are approval/request waterfall listeners. Return an outcome to answer for an owned agent or call next() to delegate. Agent-scoped listeners receive only that agent's requests; compose one terminal answerer per deployment because sibling listener order is not a policy priority mechanism. The ACP automation bridge supplies one-shot machine decisions for sessions it owns.
ApprovalPolicy is 'ask' or 'never'. The effective value is the last approval/policy event, falling back to config; setApprovalPolicy() is the write path. 'never' rejects before interactive dispatch and is the only policy stated in the prompt. Switches produce at most one coalesced notice, attributed to the user when the override follows the last request/header and to operator/config otherwise.
The tools pipeline routes ask decisions through this seam and fails closed when it is absent; the sandboxed bash tool also uses it for escalated retries. The ACP automation bridge answers calls for its own agents through the client's machine policy. Audit events remain log-only, so the model sees only the asking consumer's result. See the approval-seam Agent Note and sandbox Agent Note.
Model Experience
System prompt and policy notice
What the model sees
Under ask, every agent request carries the ask-policy prompt section below. Under never, it carries the never-policy prompt section below. A policy switch injects exactly The approval policy changed from "<old>" to "<new>" (changed by the user). or The approval policy changed from "<old>" to "<new>" (changed by the operator/config). before the next step.
Ask-policy prompt section
<!-- dsh-user-approval-policy:ask -->
Never-policy prompt section
Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).
<!-- dsh-user-approval-policy:never -->
Token effect
Small fixed per-request cost, larger under never; a change notice is conditional and retained in history.
KV Cache effect
Prefix-stable while the approval policy is unchanged. An ask/never switch changes the system-prompt section and invalidates reuse from its first changed token; the accompanying notice is append-only.
Tool outcome
What the model sees
approval/asked and approval/decided are log-only. The model sees only the asking consumer's eventual allowed, rejected, cancelled, or unavailable tool outcome; the human permission UI is not context.
Token effect
Zero duplicate audit tokens. A rejection may replace a normal tool result with a small retained error, while an allowance leaves the consumer's ordinary result.
KV Cache effect
Append-only; newly visible content follows the reusable request prefix and does not invalidate existing KV-cache entries.
Known Limitations and Deferred Work
- Requests are valid only inside an open turn — an idle or between-turn caller throws before auditing; a durable out-of-turn approval workflow is deferred.
- Only one-shot grants exist — the outcome vocabulary has
allowed-oncebut noallow-always, remembered rule, revocation, or grant store; session policy is onlyask/never. - The request carries no tool arguments — an answerer sees the tool name, reason, and optional call id; the ACP machine channel requires a call id and delegates requests without one.
- No built-in answerer — headless or incompletely composed deployments resolve
unavailableand fail closed; the service itself never prompts a human.