Skip to content

docs(rfc): sandbox seals the gateway checkout's git control files - #25692

Merged
kyleseaman merged 1 commit into
mainfrom
docs/rfc-sandbox-git-readonly
Oct 29, 2026
Merged

kyleseaman merged 1 commit into
mainfrom
docs/rfc-sandbox-git-readonly

Conversation

@iamwhatever

@iamwhatever iamwhatever commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Record the decision to seal the gateway checkout's git control files inside the agent sandbox, so #24812 has an accepted RFC on main.

PR #24812 makes the gateway checkout's .git config, hooks and info read-only inside the sandbox. First Principles blocks it as a product-shape change with no recorded decision: agents lose git push -u, git remote add, gh pr checkout and pre-commit install in that checkout. This PR is that RFC.

Why it matters

Without a recorded decision, #24812 stays blocked, and a later reader cannot tell whether the lost commands were chosen or accidental.

Not a goal

What changed (motivation → approach → change)

One new RFC, docs/request-for-change/rfc-sandbox-gateway-checkout-git-seal.md (status accepted), and its row in the docs/request-for-change/README.md index. It states the threat, the decision, what agents lose and where, the workaround, per-OS behavior, known limits (the Linux commondir gap leaves a person's terminal git exposed), alternatives and rollback.

Scope was read from #24812's code at ee9bb8e5f5: gateway_checkout_roots() in sandbox_git_control.py covers only KIROCREW_PROJECT_DIR and the gateway's own source checkout. Linked worktrees of that checkout share its sealed config, so the RFC names a separate clone as the workaround. Other repositories are unaffected.

The accepted status takes effect when a maintainer merges this PR. A maintainer who does not accept it should set status: draft instead.

Backwards compatibility

Compatible: documentation only.

Removes nothing: this PR adds one document and one index row.

Tests

N/A — docs only. scripts/docs-lint.sh passes.

Manual verification

N/A — docs only; claims about main were read at 7df9837c60, claims about the change at #24812 head ee9bb8e5f5.

Related Issues

Refs #24812

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

@iamwhatever
iamwhatever requested review from a team and KiroCrewCaptain as code owners October 29, 2026 07:10
@iamwhatever
iamwhatever requested a review from CrysisDeu October 29, 2026 07:10
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 29, 2026
@github-actions

github-actions Bot commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ⏭️ skipped

Revision 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Does not block merge.

@github-actions

github-actions Bot commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

This PR only adds a document. It records a sandbox-widening decision for a maintainer to approve, which is the route AGENTS.md requires.

Notes

  • Section 1 calls fix(sandbox): write-protect git control files in the gateway's checkout #24812 "the second layer" for a person's terminal git. Section 6.1 says that on Linux, the agent can still create a missing commondir and point it at its own config and hooks. So on Linux the terminal-git threat stays open. A later reader could take the loss of push -u and gh pr checkout as buying protection that it doesn't fully buy there. The RFC should state this trade in its Decision section.

[DESIGN-REVIEWED] 5cfa2bb

Description read

[DESCRIPTION-READ] a566617afaf780e66651a0c04751e06076ef4a1cbb1e190de4dc1288349a4fcb

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 5cfa2bb

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 5cfa2bb

Verdict parsed from the review's SHA-scoped output markers for commit 5cfa2bbb1c8b4df4128b63af361020a7368d1e77.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 5cfa2bbb1c8b4df4128b63af361020a7368d1e77 false-positive: <reason, 20 to 500 characters>

@iamwhatever
iamwhatever force-pushed the docs/rfc-sandbox-git-readonly branch from ce534fd to 020ea27 Compare October 29, 2026 07:28
@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 29, 2026
@iamwhatever
iamwhatever force-pushed the docs/rfc-sandbox-git-readonly branch from 020ea27 to 5cfa2bb Compare October 29, 2026 14:51
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 29, 2026
@kyleseaman
kyleseaman merged commit 6c02136 into main Oct 29, 2026
94 checks passed
@kyleseaman
kyleseaman deleted the docs/rfc-sandbox-git-readonly branch October 29, 2026 16:27
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 29, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants