A shallow clone leaves a cited commit unchecked by pauldambra · Pull Request #86 · PostHog/coherence

pauldambra · 2026-10-09T22:20:26Z

🤖 This PR was written by a robot (Claude, via PostHog Desktop) for @pauldambra.

Why

In a shallow clone, the Stop hook said that correct commit citations in practice files were "no commit in this repository". A hosted agent sandbox clones at depth 1, so every older commit looks missing. In ops-platform, all the cited commits exist on GitHub.

What

missingCommits now asks git if the clone is shallow. It asks only when some cited commit is missing. In a shallow clone, a commit that is not there is not reported, because the clone depth may have cut it off.
A full clone still reports a commit that does not exist. The workflow that ci install writes fetches full history, so CI still catches a wrong citation.
coherence defect --introduced <sha> had the same problem: in a shallow clone it refused any older commit. The check now lives in one place, src/journal/commits.ts, and both the practice check and defect --introduced use it.
In a shallow clone, a hash that names a file or tree the clone holds, or more than one object, is still reported, because it is no commit at any depth.
New invariant in Spec.spec.md and parameterized tests in practice.test.ts and defects.test.ts, and a new invariant in Journal.spec.md.
Bumps the version to 1.7.1. When this merges, the Release workflow publishes 1.7.1 to npm.

Checks

On the depth-1 ops-platform clone, spec --check gave 12 "no commit" problems before this change and 0 after.
typecheck, spec:check and lexicon:check pass.
practice.test.ts and defects.test.ts pass.
Both new invariants were refuted with coherence refute (each break made its test go red), restored, and found green by run, batched and alone. Both read as invariant.

🤖 Generated with Claude Code

Created with PostHog Desktop

A practice's cited commit was a problem whenever git cat-file said missing. In a depth-1 clone, such as a hosted agent sandbox, every older commit is missing, so Stop reported correct citations as "no commit in this repository". When some cited commit is missing, git is now asked if the clone is shallow; if it is, the citations are left unchecked. A full clone, as the workflow ci install writes fetches, still catches a wrong one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

posthog · 2026-10-09T22:20:32Z

🦔 PostHog Review (standard) reviewed this pull request

Nothing worth raising.

chatgpt-codex-connector · 2026-10-09T22:21:06Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T22:23:14.502937Z	`00f1d8d`	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 00f1d8d0f0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

pauldambra

Note

🤖 Automated comment by QA Swarm — not written by a human

QA Swarm review complete. See inline comments.

pauldambra · 2026-10-09T22:28:36Z

Note

🤖 Automated comment by QA Swarm — not written by a human

Multi-perspective review: router (cheap-first pass) + delegated lenses (qa-team, paul-reviewer, xp-reviewer, security-audit, engineering-systems-thinking as warranted)

Verdict: ✅ APPROVE _{(round 3 @ ae1534c)}

Moving the shallow-aware check into src/journal/commits.ts keeps the old behaviour of defect --introduced in a full clone, outside a git checkout, and for an ambiguous short hash. It also fixes the same shallow-clone problem there.

Key findings

⚪ NIT src/spec/practices.ts:111: double blank line left after the helpers moved out.

Reviewer summaries

Reviewer	Assessment
🧭 router (sonnet)	Danger LOW, confidence HIGH, no delegation. Checked the old and new git calls against real git. Unit tests not run in the sandbox; CI ran them on `ae1534c` and all 16 checks passed.

Previous rounds (2)

round 1 @ 00f1d8d — 💬 APPROVE WITH NITS: shallow fallback hid provable wrong citations (fixed in c3ad7bb); refuted: line missing (left for a human).
round 2 @ c3ad7bb — ✅ APPROVE: no new findings.

Automated by QA Swarm — not a human review

The shallow fallback left every commit citation unchecked once one was missing. But a hash that names a file or a tree the clone holds, or several objects, is no commit at any depth. In a shallow clone, the missing citations are now asked of git again without ^{commit}, and only those that name nothing here stay unchecked. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

defect --introduced asked git cat-file -e <sha>^{commit}, so in a depth-1 clone it refused any older commit as "no commit by that name", the same blind spot the practice citation check had. The shallow-aware check moves from spec/practices.ts to journal/commits.ts, and both the practice check and defect --introduced use it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

pauldambra

Note

🤖 Automated comment by QA Swarm — not written by a human

QA Swarm review complete. See inline comments.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

…by via The parameterized tests gave each case its own name, so no test was named by the via: lines, and run and refute selected nothing. Each loop now sits in a describe block named by its via. Both invariants were then refuted with the break staged, restored byte for byte, and run green batched and alone; the run store records it. The alone run used NODE_OPTIONS=--test-reporter=tap, because Node 24 prints "ℹ pass" by default and the config's testMatch reads "# pass". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

A shallow clone no longer reads a correct commit citation as no commit: practice citations and defect --introduced ask git whether the clone is shallow, and there report only a hash that names some other object, or several (#86). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

pauldambra self-assigned this Oct 9, 2026

pauldambra marked this pull request as ready for review October 9, 2026 22:20

chatgpt-codex-connector Bot reviewed Oct 9, 2026

View reviewed changes

Comment thread src/spec/practices.ts Outdated

pauldambra commented Oct 9, 2026

View reviewed changes

Comment thread src/spec/practices.ts Outdated

Comment thread src/spec/Spec.spec.md

pauldambra added 2 commits October 9, 2026 23:29

pauldambra commented Oct 9, 2026

View reviewed changes

Comment thread src/spec/practices.ts

pauldambra added 3 commits October 10, 2026 00:02

Remove a double blank line left by the moved helpers

b8899dd

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Generated-By: PostHog Desktop Task-Id: b53cdddb-2e4d-43d2-97ec-a3b067257f29

pauldambra merged commit a73b33a into main Oct 10, 2026
16 checks passed

pauldambra deleted the posthog/shallow-clone-cited-commits branch October 10, 2026 10:16

Navigation Menu

Repository navigation

A shallow clone leaves a cited commit unchecked - #86

A shallow clone leaves a cited commit unchecked#86
pauldambra merged 6 commits into
mainfrom
posthog/shallow-clone-cited-commits

pauldambra commented Oct 9, 2026 •

edited

Loading

Uh oh!

posthog Bot commented Oct 9, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

pauldambra left a comment

Uh oh!

Uh oh!

Uh oh!

pauldambra commented Oct 9, 2026 •

edited

Loading

Uh oh!

pauldambra left a comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pauldambra commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why

What

Checks

Uh oh!

posthog Bot commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦔 PostHog Review (standard) reviewed this pull request

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

pauldambra left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

pauldambra commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Verdict: ✅ APPROVE (round 3 @ ae1534c)

Key findings

Reviewer summaries

Uh oh!

pauldambra left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

pauldambra commented Oct 9, 2026 •

edited

Loading

posthog Bot commented Oct 9, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 9, 2026 •

edited

Loading

pauldambra commented Oct 9, 2026 •

edited

Loading

Verdict: ✅ APPROVE _{(round 3 @ ae1534c)}