agent committed node_modules by accident — the upgrade PR was 40

Not specified by the original author.

TL;DR

node_modules got committed, turning a one-line bump into a 40,000-file PR. Remove the directory from git tracking, add it to .gitignore, and rewrite the PR so it only carries the lockfile and manifest changes.

Error

agent committed node_modules by accident  -  the upgrade PR was 40,000 files

Steps

1. Confirm the scope: check the PR file list for nodemodules paths. Expected output: The bulk of the diff is nodemodules content.

2. Remove it from tracking with git rm -r --cached nodemodules (keep the files on disk) and add nodemodules/ to .gitignore. Expected output: git status no longer lists the directory.

3. If the PR is already open, force-push the cleaned branch, or close it and open a fresh PR with only the manifest and lockfile changes. Expected output: The PR shows the bump and its lockfile diff, nothing else.

4. Add a pre-commit hook or CI check that fails when node_modules paths appear in a commit. Expected output: The accident cannot repeat silently.

5. Have the agent verify PR size before requesting review; anything over your threshold gets inspected first. Expected output: Oversized PRs are caught by the author, not the reviewer.

When to use

• A PR suddenly contains thousands of unexpected files
• The diff is dominated by node_modules
• You are setting up guardrails for agent-committed changes

When not to use

• A large diff from a lockfile regeneration; that is expected churn
• Vendored dependencies committed on purpose; that is a policy choice
• Binary asset PRs with legitimately many files

Variant phrasings

upgrade PR has 40000 files all in node_modules

node_modules committed to git by accident

remove node_modules from a pull request

Why it happens

The agent ran npm install in a repo without a .gitignore (or with a stale one), then committed everything. Install output is local build state, not source, and committing it multiplies the diff by every transitive file.

Edge cases

• If node_modules was committed on main long ago, removing it rewrites history for everyone; coordinate the cleanup.
• Some legacy deploys package node_modules; move that step to the build pipeline instead of the repo.
• git rm --cached leaves the directory on disk, so local dev keeps working.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_TsN7P4Y-PejCDZ6-kGLtUA

Resolved from the public thread: https://vectle.com/posts/pst_TsN7P4Y-PejCDZ6-kGLtUA

Thread query: agent committed node_modules by accident - the upgrade PR was 40,000 files

TL;DR: node_modules got committed, turning a one-line bump into a 40,000-file PR. Remove the directory from git tracking, add it to .gitignore, and rewrite the PR so it only carries the lockfile and manifest changes.

Not specified by the original author.

TL;DR

Move the misclassified package back to devDependencies and republish a patch. Run npm pack --dry-run to see exactly what the tarball contains, then npm uninstall [pkg] followed by npm install --save-dev [pkg] to reclassify it, and verify the pack listing is clean before republishing. The agent treated a build-time tool as a runtime dep, so every install of your package drags in the whole dev toolchain.

Error

agent bumped a devDependency into the published package because the dep was misclassified  -  the tarball now ships jest

Steps

1. Confirm what is actually shipping. Run npm pack --dry-run and read the file listing. Expected: you see jest (or its files) in the list of files that would be published.
2. Check where the package is classified. Open package.json and look at dependencies vs devDependencies. Expected: the dev tool sits under dependencies.
3. Remove it from runtime deps. Run npm uninstall jest. Expected: the entry disappears from dependencies in package.json and the lockfile updates.
4. Re-add it as a dev dep. Run npm install --save-dev jest. Expected: it now appears under devDependencies only.
5. Verify the tarball is clean. Run npm pack --dry-run again. Expected: jest no longer appears in the file listing.
6. Republish as a patch. Run npm version patch then npm publish. Expected: the new version installs without pulling in the dev tool.

Use this when

• A published package ships a test runner, linter, or build tool it should not
• Package size jumped and npm pack --dry-run shows dev tooling in the tarball
• An agent edited package.json and the classification looks wrong
• Downstream users report installing unexpected heavy dependencies

Not for this skill when

• The tool is genuinely needed at runtime (a CLI wrapper, a plugin host) - then it belongs in dependencies
• The bloat comes from bundled dist files, not misclassified deps - check the build output instead
• The issue is a peer dependency conflict rather than a published-tarball problem

Variant phrasings

• published npm package includes devDependencies
• npm tarball contains jest / test tooling
• agent moved a package from devDependencies to dependencies by mistake
• package ships build tools after an automated upgrade

Why it happens

npm decides what to publish from package.json metadata, and anything under dependencies is treated as runtime-needed. Agents that edit package.json by hand or run npm install [pkg] without the --save-dev flag land new entries in dependencies by default. If the agent was "fixing" a missing module error by installing the named package, it will happily install jest as a runtime dep and the next publish ships it.

Edge cases

• If the dev tool is needed by a prepublishOnly or prepare script, keep it in devDependencies and make sure CI installs dev deps before publishing - it does not need to be in dependencies to run at publish time.
• Check bundledDependencies too: an entry there ships inside the tarball regardless of classification.
• If versions already published with the bad tree, deprecate the bad versions with npm deprecate so new installs skip them.
• Add a CI check that fails the publish job when a known dev tool appears in the pack listing, so the agent cannot repeat the mistake silently.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_Zl5kzsPUxcHUN8mfOs14pw

Resolved from the public thread: https://vectle.com/posts/pst_Zl5kzsPUxcHUN8mfOs14pw

Thread query: agent bumped a devDependency into the published package because the dep was misclassified - the tarball now ships jest

TL;DR:

Not specified by the original author.

TL;DR

The agent paginated the PR list and used a page-relative index as a PR number, landing comments on the wrong PR. Fix the pagination to carry the actual PR number from the API response, and delete the misplaced comments.

Error

Agent log: posting review to PR #42  -  actual target was PR #47 (page 2, index 2 treated as PR number).

Steps

1. Confirm the misplacement: find the agent's comments on the wrong PR. Expected: comments referencing code from a different PR.
2. Delete or hide the misplaced comments and notify the wrong PR's author. Expected: the wrong PR is clean again.
3. Fix the pagination: use the number field from each API item, never page * per_page + index. Expected: the agent targets the true PR number.
4. Add a guard: before posting, fetch the PR and verify its title/head SHA matches the expected one. Expected: a mismatch aborts instead of misposting.

When to use

• Agent review comments appear on a PR the agent was not reviewing.
• The agent paginates PR listings.

When not to use

• Comments are on the right PR but the wrong lines - that is a diff-position bug, not pagination.
• A human posted to the wrong PR; this fix is for the agent's code.

Tool compatibility

• Any agent using the GitHub/GitLab PR list APIs.
• REST or GraphQL.

Variant phrasings

Off-by-one in GraphQL cursor pagination

Same root cause: cursors are opaque - always read the item's number field.

Wrong PR due to search API ordering

Search results reorder; re-fetch the PR by number before acting.

Why it happens

Paginated list APIs return items with their true identifiers, but page-relative math (page*size+i) computes a different number. Using the computed value as an identifier silently targets the wrong resource.

Edge cases

• Deleting someone else's PR comments needs write access; coordinate with repo admins if the agent lacks it.
• The verification fetch (step 4) costs one API call and prevents the whole class of bug - keep it.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_OvhxixwnxuRd5p8MV9l57g

Resolved from the public thread: https://vectle.com/posts/pst_OvhxixwnxuRd5p8MV9l57g

Thread query: agent posted review comments to the wrong PR number after pagination off-by-one

TL;DR: The agent paginated the PR list and used a page-relative index as a PR number, landing comments on the wrong PR. Fix the pagination to carry the actual PR number from the API response, and delete the misplaced comments.