fix: prevent cross-PR push from subagent-induced branch switch (#796)

* fix: prevent cross-PR push from subagent-induced branch switch

A workflow_dispatch run for zed-industries/cloud (workflow run 26036155393)
force-pushed the orchestrator's work onto an unrelated engineer's PR branch
(origin/reactivate-pro-plan, PR #2582). The orchestrator's reviewfrog subagent
called checkout_pr({pull_number: 2582}), which (1) moved the shared working
tree to pr-2582 and (2) persisted pushDest pointing at reactivate-pro-plan.
The orchestrator's subsequent commit + push_branch then clobbered the victim
PR. Recovery + disclosure in PR #2584.

Three compounding bugs closed here:

1. checkout_pr dirty-tree guard had a first-call hole: the previous condition
   required ctx.toolState.issueNumber to already be set, so on workflow_dispatch
   runs the first checkout_pr (commonly from a subagent) bypassed the guard
   entirely. Now any PR switch with a dirty tree is refused, including the
   first switch of a run. Idempotent same-PR re-checkouts are still absorbed
   by alreadyOnBranch inside checkoutPrBranch.

2. push_branch trusted sticky pushDest blindly. Added a backstop: refuse
   pushes where the local branch matches /^pr-(\d+)$/ AND pushDest.remoteBranch
   differs from it AND the current run is not scoped to PR N (event.is_pr === true
   && event.issue_number === N). This catches subagent-induced silent branch
   switches even if a future bug reintroduces a first-call hole in fix #1.

3. Build-mode self-review prompt told the orchestrator to ship "the output of
   git diff" to the reviewer. The model in this run synthesized
   git diff main...HEAD, which excludes uncommitted work — and Build self-review
   runs BEFORE the commit, so the reviewer saw an empty diff and thrashed,
   eventually calling checkout_pr on a random PR to find something to look at.
   Prompt now specifies git diff origin/<base-branch> (two-dot, no HEAD),
   which compares the working tree against the remote base.

Refs:
  zed-industries/cloud workflow run 26036155393
  zed-industries/cloud#2582 (victim)
  zed-industries/cloud#2584 (disclosure)

* review: key dirty-tree guard on current branch + drop 'two-dot' misnomer

Address review feedback on PR #796.

1. checkout_pr dirty-tree guard now keys off the live current branch
   (git rev-parse --abbrev-ref HEAD), not ctx.toolState.issueNumber.
   issueNumber is ALSO set by get_issue / get_issue_comments /
   get_issue_events, so a subagent doing get_issue(N) followed by
   checkout_pr(N) on a dirty tree would have bypassed the original guard
   (issueNumber === pull_number). The current branch is the actual
   primitive for "would this call move HEAD" — querying it directly avoids
   correlating on toolState that other tools write to.

2. modes.ts: drop the wrong "two-dot" label on git diff origin/<base>.
   That's the single-rev form, not two-dot. Copilot was right that the
   label was confusing/contradictory with the actually-shown command.
This commit is contained in:
Colin McDonnell
2026-05-20 05:24:24 +00:00
committed by pullfrog[bot]
parent a0576a702a
commit 6c166ac1cc
3 changed files with 31 additions and 11 deletions
+8 -10
View File
@@ -805,16 +805,14 @@ export function CheckoutPrTool(ctx: ToolContext) {
return inFlight;
}
// refuse to clobber an active checkout if the working tree is dirty —
// forces the agent to commit/push or discard before switching contexts
// instead of silently overwriting uncommitted work. `issueNumber`
// tracks any issue/PR the agent has touched (issues and PRs share
// GitHub's number space); the guard fires only when the agent is
// switching to a *different* number with a dirty tree, which captures
// the legitimate "stop, you have unsaved work" case regardless of
// whether the prior number was an issue or a PR.
const current = ctx.toolState.issueNumber;
if (current !== undefined && current !== pull_number) {
// refuse to clobber an uncommitted tree whenever this call would move
// HEAD away from the target pr-N branch. keyed off the live current
// branch (not toolState.issueNumber, which is also written by
// get_issue / get_issue_comments / get_issue_events and so doesn't
// mean "currently checked out"). catches the subagent-sharing-cwd
// case from zed-industries/cloud (2026-05-18).
const currentBranch = $("git", ["rev-parse", "--abbrev-ref", "HEAD"], { log: false }).trim();
if (currentBranch !== `pr-${pull_number}`) {
const dirty = $("git", ["status", "--porcelain"], { log: false }).trim();
if (dirty) {
throw new Error(
+22
View File
@@ -253,6 +253,28 @@ export function PushBranchTool(ctx: ToolContext) {
// validate push destination matches expected URL
const pushDest = validatePushDestination(ctx, branch);
// backstop against subagent-induced cross-PR clobbers: a subagent
// shares cwd + toolState with the orchestrator, so its `checkout_pr(N)`
// moves HEAD to pr-N and persists pushDest pointing at the foreign
// PR's remote branch. refuse pr-N → origin/<other> pushes unless this
// run is itself scoped to PR N (zed-industries/cloud, 2026-05-18).
const prBranchMatch = branch.match(/^pr-(\d+)$/);
if (prBranchMatch && pushDest.remoteBranch !== branch) {
const prNumber = Number(prBranchMatch[1]);
const event = ctx.payload.event;
const runScoped = event.is_pr === true && event.issue_number === prNumber;
if (!runScoped) {
throw new Error(
`push blocked: local branch '${branch}' would push to '${pushDest.remoteName}/${pushDest.remoteBranch}', ` +
`but this run is not scoped to PR #${prNumber}. ` +
`the 'pr-${prNumber}' branch was created by a prior checkout_pr call (likely from a subagent — subagents share the working tree and toolState with the orchestrator). ` +
`you have probably landed your commit on the wrong branch. ` +
`switch to your own feature branch first (e.g. 'git checkout <feature-branch>') and then push. ` +
`if the push to PR #${prNumber} is intentional, this run needs to be triggered against that PR.`
);
}
}
// block pushes to default branch in restricted mode
if (pushPermission === "restricted" && pushDest.remoteBranch === defaultBranch) {
throw new Error(