Repository navigation
Conversation
This was referenced Nov 15, 2026
Restack the reviewed #1327 reply approval feature (tree delta 853e862..35dcb34, previously at 88703be) onto main after the Agent package extraction (#1343), without the #1337 extraction spike. Agent owns the conversation, reply review, acceptance record and replay (packages/agent/src/reply-acceptance.ts, reply-review.ts, exported from @executablemd/agent/host). Core keeps the generic pieces the feature needs: the accepted-reply copy grants, reply secret location and text checks, Elicit answer readers, and executeInstalledBound with the evaluator that takes an AcceptedRootReply, all exported from @executablemd/core/host. Core imports nothing from the Agent package. The approval drawer takes precedence over main's Sidekick inspection view, and the chat projection keeps both main's request reading and the feature's control outcomes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
taras
force-pushed
the
codex/sidekick-reply-approval-correction
branch
from
November 16, 2026 05:05
88703be to
737601c
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
When secret detection flags a completed Sidekick reply, the REPL can already preserve the conversation by withholding the reply. This change lets the person inspect the finding and decide whether to keep the original reply and continue.
Refs #1311.
The feature is restacked on main at
8c71b35b6after the Agent extraction in #1343.What changes
A flagged reply opens a foreground approval drawer. It shows the detection rule and position and masks the match until the person chooses Reveal. Withhold, closing the drawer, and stopping the turn preserve the withheld completion. Accept saves the exact original reply, resumes the same turn, and permits ordinary ACP output and automatic feedback.
How it works
Review guide
Start with the “Review a withheld Sidekick reply” section in
specs/repl-spec.mdand the “Sidekick reply approval” mounted suite.packages/agent/src/reply-review.tsandpackages/agent/src/conversation.ts.packages/agent/src/reply-acceptance.ts.packages/cli/src/repl/sidekick/control-outcomes.tsandpackages/core/src/secrets/reply-grant.ts.packages/cli/src/repl/application.tsandpackages/cli/src/repl/program.ts.What must stay true
How to verify it
deno task test --filter 'Tier CW' packages/agent/tests/agent-conversation.test.tsexercises the continuation, journal, and scoped secret policy.737601c97, including the 14 reply-approval cases and the feedback acknowledgement-loss regression. It covers same-turn acceptance through real ACP, protected questions, TODO feedback, and Reveal at three terminal sizes.xmd repl#1311 recovery suite, Deno typecheck, and JSR dry run also passed.737601c97. The original independent Fable reviewer re-reviewed the same commit and returned PASS.New abstractions
Risks and limitations
Scope confirmation
Visual verification
Both clean builds received the same
Check the reply.prompt and synthetic credential-like reply through the mounted REPL harness. The real scanner and renderer handled both turns. The 120×30 and 72×20 frames were captured after resizing from 160×36 at the withholding checkpoint.853e86268353937db2f746e47e3824b2a48c5f5011f3c8a70093454d25bb55501fdc9d3f97383259The base shows that it withheld the reply. The PR head identifies the rule and position, masks the matched value, and offers Reveal, Withhold, and Accept. At 72×20, scroll the drawer to reach the decision controls.
The mounted test accepted the reply at 160×36 and saw the original output appear in the same turn. It withheld the reply at 120×30 and 72×20 and verified the output stayed withheld. Each decision ran with one provider prompt.
REPL typing latency
One local paired run on macOS arm64 with Deno 2.9.7, using the same pinned base and head. The benchmark measured from a keystroke until the new composer text appeared at 200×52. Each workload ran 40 edits, with the first 10 excluded as warm-up.
Median latency changed by under 0.2 ms on both workloads. The small-history p95 rose 0.56 ms, and the wrapping-history p95 fell 0.54 ms in this run. The four focused typing regression tests also passed on the PR head.