fix(disk-hygiene): do not flag PowerShell 2>&1 as file redirect - #2627
Merged
Conversation
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
Contributor
|
Claude finished @kyle-sexton's task in 1m 27s —— View job Security review — no issues found
This PR changes the regex Old: Analysis:
No injection, authorization, secrets-handling, or CI/workflow-permission issues in this changeset — it's a self-contained Python regex fix plus a test and changelog/version bump. No security concerns to report. |
kyle-sexton
enabled auto-merge (squash)
August 14, 2026 22:49
|
Last security-reviewed head: |
Exclude stream-merge forms (`2>&1`, `*>&1`) from `_POWERSHELL_OUTPUT_REDIRECT` so ordinary stderr capture no longer prompts as a file-overwriting mutation (#2615). File redirects like `2>out.txt` still ask. Co-authored-by: Kyle Sexton <kyle-sexton@users.noreply.github.com>
cursor
Bot
force-pushed
the
cursor/fix-disk-hygiene-ps-stream-merge-f8b0
branch
from
August 14, 2026 22:57
45d24d5 to
1502bd8
Compare
Contributor
|
Claude finished @cursor[bot]'s task in 26s —— View job Code review — skipped
Skip gate: this PR is already MERGED (not open), which matches skip-gate condition 1 ("PR is closed or not open") in the For reference, this PR was already reviewed for security by the |
|
Claude has reviewed this PR 1 time. The lane skips further automatic reviews after 5; deleting this comment resets the count. |
kyle-sexton
added a commit
that referenced
this pull request
Aug 15, 2026
…direction (#2671) ## Problem PR #2627 closed issue #2615 by excluding stream merges (`2>&1`, `1>&2`, `*>&1`) from `_POWERSHELL_OUTPUT_REDIRECT`. That fix is correct but partial: it only excludes `&`. In a **discard** the character after `>` is `$`, not `&`: ```python # on origin/main _POWERSHELL_OUTPUT_REDIRECT = re.compile(r"(?<![<>])>(?![=>&])") ``` So `2>$null` — PowerShell's `/dev/null`, and the standard way to silence a noisy read-only command — still matches, and the guard still prompts with *"disk-hygiene flagged shell output redirection (may overwrite a file)"*. The issue closed while the operator's actual failing commands kept prompting. ## Observed cost Two read-only commands from an attended session, both prompted: - `gh issue list --repo <owner/repo> --state open 2>$null` - `chezmoi managed | Select-String -Pattern claude 2>$null` Neither can write a file. Each prompt stalled the session. This is the approval-fatigue mechanism issue #2615 itself describes: a destructive-action guard that cries wolf on read-only work measurably degrades the signal of its genuine prompts. ## Fix ```python _POWERSHELL_OUTPUT_REDIRECT = re.compile( r"(?i)(?<![<>])>(?![=>&])(?![^\S\n]*\$null(?![\w-]))" ) ``` Three deliberate details: - **`(?i)`** — PowerShell variable names are case-insensitive, so `2>$NULL` is the same discard. - **`[^\S\n]*`** — only *horizontal* whitespace is skipped, so a trailing `>` at end of line cannot borrow a `$null` from the next line. - **`(?![\w-])`** — `$nullish` and `$null-backup` are ordinary variables, not the null device, and must still be treated as file targets. The exclusion is spelled the way guardrails' `ps::write_bypass` already spells the same `$null` exclusion, rather than inventing a second spelling for the same concept in the same fleet. ## Coverage Allowed (newly, 9 forms): `2>$null`, `*>$null`, `>$null`, `2> $null`, `2>$NULL`, `2>$null; <cmd>`, `2>$null | <cmd>`, and the two real-world commands above. Still flagged (7 forms): `2>out.txt`, `> out.txt`, `1>file`, `2>$nullish`, `2>$null-backup`, `2>$null > out.txt`, `2>&1 > out.txt` — the last two being the important ones: discarding one stream while redirecting another is still a file write, and the second `>` has no `$null` after it. ## Test evidence `test_powershell_null_discards_are_not_file_redirects` covers every form above. Full-suite comparison, same interpreter, same machine: | tree | tests | result | |---|---|---| | `origin/main` (pristine, extracted via `git archive`) | 282 | 3 failures, 4 skipped | | this branch | 283 | **the same 3 failures**, 4 skipped | The three pre-existing failures are `test_deny_emits_blocked_telemetry_when_sink_wired`, `test_preview_allows_root_children_os_managed_snapshot`, and `test_stash_must_exist_in_an_independent_checkout`. They are identical on both trees and unrelated to this change; this branch adds one passing test and introduces no new failure. ## Adjacent gap found, deliberately NOT fixed here `>>` (append) is matched by **neither** the old nor the new pattern — `(?![=>&])` rejects the first `>` of the pair, and the cmdlet word list catches `out-file`/`add-content` but not a bare `>>` redirect. So `<cmd> >> append.txt` writes a file without a prompt. That is pre-existing on `main` and orthogonal to this change, so it is reported rather than folded in. Worth its own issue. ## Related - #2627 — the stream-merge fix this completes; it closed #2615 for the `>&` form only. - #2674 — concurrent PR against the same plugin; it claims version `0.20.3`, this one `0.20.2`. - The `>>` append gap described above is unfiled and needs its own issue. Closes #2615 --------- Co-authored-by: Kyle Sexton <kyle-sexton@users.noreply.github.com>
kyle-sexton
added a commit
that referenced
this pull request
Aug 15, 2026
<!-- CURSOR_AGENT_PR_BODY_BEGIN --> Closes #2675 ## Summary `_POWERSHELL_OUTPUT_REDIRECT` matched neither character of a `>>` pair, so `<cmd> >> append.txt` wrote a file with no prompt while the same command with `>` prompted. ## Fix - Explicit append matcher (`>>` / `2>>` / `*>>`) without widening stream-merge / `$null`-discard behavior - `>> $null` stays silent, requiring a real token terminator after `$null` (same boundary as #2671) - Punctuation continuations like `>>$null/out.txt` stay flagged - Plugin version `0.20.5` (above #2671's `0.20.4`) ## Verification Focused GuardTests for append redirects and stream merges pass locally. ## Related Refs #2615 / #2627 / #2671 — stream-merge and `$null`-discard carve-outs that must stay narrow. <!-- CURSOR_AGENT_PR_BODY_END --> Co-authored-by: Kyle Sexton <kyle-sexton@users.noreply.github.com>
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
2 participants
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.
Closes #2615
Summary
The disk-hygiene PowerShell belt treated
2>&1stream merges as file-overwriting redirects because_POWERSHELL_OUTPUT_REDIRECTmatched the>in2>&1.Fix
Exclude PowerShell stream-merge redirections (
N>&M) from the file-overwrite detector. Add regression coverage.Verification
See branch CI / local destructive_guard / hygiene tests on the PR checks.
Related
Refs #2589 — argument-surface docs (serial disk-hygiene lane).
Refs #2591 / #2618 — belt lifetime (later wave).