Query and manage Linear issues, projects, teams, cycles, labels, and comments through a managed OAuth GraphQL endpoint.
Data & analysis
Ward Public Repo Git Gate
Try itRun the public-repository gates for Git commits, pushes, and pull requests. Check staged and untracked scope, secrets and personal data, internal material, branches and remotes, PR base/head, CI, and merge prerequisites. Use when the user asks to commit, push, open a PR, review public repository content, or run the complete public-repository gate.
What it does
Treat Git commits, remote pushes, and pull requests as sequential public-content gates. is not a security boundary; inspect what Git will actually commit or has already put under version control.
The skill document
Public Repository Git Gate
Treat Git commits, remote pushes, and pull requests as sequential public-content
gates. .gitignore is not a security boundary; inspect what Git will actually
commit or has already put under version control.
Keep checks read-only by default. Passing a check does not authorize a commit, push, PR creation, or remote-branch update. The user must explicitly request those actions, which must also follow the repository's branch and approval rules.
Run the script
Run the script from this skill directory. The three checks are not repeated full scans: commit checks inspect the index, push checks inspect changes from the base, and CI independently verifies the final PR head.
python3 scripts/public_repo_check.py --repo /path/to/repository --staged --require-feature-branch
Run the staged check before committing and the change/remote check before pushing:
python3 scripts/public_repo_check.py --repo /path/to/repository --staged --require-feature-branch
python3 scripts/public_repo_check.py --repo /path/to/repository --changed-since origin/main --check-remote
Use --all for a full baseline when introducing the gate, changing its rule
configuration, recovering from an accidental push, or preparing a Release.
Projects can add scoped exceptions and rules with
--config path/to/public-repo-gate.json. Keep the default rules conservative;
document each intentionally public special resource in project configuration
instead of disabling the entire check.
For a large repository, replace origin/main with the PR's exact base SHA. Do
not use an old local base. If the ref is missing, fail the check and synchronize
the base first rather than assuming that there are no changes.
Layered check strategy
| Stage | Scope | Checks | Target duration |
|---|---|---|---|
| commit | staged files | Paths, secrets, diff format, feature branch | Seconds, offline |
| push | base...HEAD changes | Incremental public-content scan and remote/branch/upstream checks | Seconds; no full test suite |
| PR | final head | Full CI security scan, lint, test, build, and PR diff review | CI-owned |
| Release | final artifacts and live state | Archives, licenses, tag, deployment, and attachments | Separate gate |
A passing commit check cannot replace remote push/PR confirmation, and an incremental push check cannot replace PR CI. Each layer should cover only its own responsibility.
Commit gate
- Confirm the repository visibility first. Treat the remote repository, branch, PR, Preview, and history as public content by default.
- Check
git status --short --branch, the current branch, and the target PR base. Do not commit ordinary changes directly tomainormaster. - Stage only the necessary files. Inspect staged names, status, and the full diff, not just a summary.
- Run the script's
--stagedcheck andgit diff --cached --check. This is a fast gate; do not require the full test suite for every commit. - For higher-risk code, run targeted tests required by the project. Let PR CI own complete lint, test, build, security, and license checks.
- Confirm that the commit message, version files, public documentation, tests, and configuration are synchronized. Keep internal plans, legal opinions, business strategy, customer data, and private agent configuration out of public commits.
Push gate
- Reconfirm the remote URL, repository visibility, target branch, upstream, and PR head before pushing. Do not infer remote state from memory.
- Run
--changed-since origin/main --check-remote, then inspect recent commits on the branch for content that should not be public. The script cannot prove the hosting platform's repository visibility; confirm that setting manually. - In the ordinary workflow, push only a feature branch, never
maindirectly. Do not use--forceto overwrite someone else's branch. History cleanup is an exception requiring a local backup and an explicit recorded reason. - After pushing, confirm through the remote platform that the branch exists, the commit SHA matches, the PR base/head is correct, and no other PR was accidentally created or updated.
- If an accidental push is found, stop pushing immediately. Preserve a backup reference, pause or close the PR, rotate exposed secrets, rebuild from a clean public base, and follow the hosting platform's process for historical objects and caches.
PR gate
- Set
mainas the PR base and the feature branch as the head. Describe the scope, risks, verification results, and unfinished work in the title and body. - Recheck the PR file list and full diff, especially new files, generated artifacts, hidden directories, configuration, licenses, and documentation.
- Require CI to pass with a full public-content scan of the final head and the project's lint, test, build, security, and license checks. Add real environment evidence for database, deployment, third-party-resource, or release changes. Passing CI does not prove that a production migration ran.
- List incremental local checks, full CI checks, project tests/build, security checks, and license checks in the PR body. Explain every failed or skipped check.
- Do not treat the author's own PR as an independent review. Obtain at least one appropriate review and request security or legal review when needed.
- Reconfirm before merging that no internal material, personal data, secrets,
or unnecessary public content remains. After merging, delete the feature
branch and independently verify the final
maincommit.
When a check fails
Stop the current commit, push, or merge; do not commit first and explain later. Classify the finding as a false positive, a project-specific public resource, or content that should not be public. For a false positive, add the smallest scoped project configuration and retain the reason. For a real issue, unstage the content or remove it from the commit. Treat exposed secrets, personal data, or internal material as a security incident; rewriting branch history alone is not sufficient.
Boundary with the Release gate
This skill covers commits, pushes, and PRs. Use the sibling
public-release-gate skill for tags, final installers, archives, GitHub
Release attachments, and production-deployment acceptance. Only both gates
together constitute a complete release review. Do not repeat the Release
artifact scan at every commit, push, and PR layer.
Related skills
Stores durable facts in a categorized, plain-markdown vault on disk, alongside your agent's built-in memory.
Find why your productivity system keeps failing, then apply the smallest fix — capacity math, bottleneck routing, durable local notes.
Run Git operations — commits, branches, merges, rebases, conflict resolution, and recovery — with safety rules enforced.
Manage Stripe customers, subscriptions, invoices, products, prices, and payments through OAuth-authenticated API calls.
Join a video meeting as an AI bot with voice, avatar, and screenshare across four operating modes.
More from wardlu
Browse all skillsAudit an AI agent's reachable data, external data paths, and untrusted instructions, then explain the result in plain language with Low, Medium, High, or Critical severity. Use when a user asks what an agent can see, where data may go, whether a skill, plugin, or tool is safe, or how to reduce agent privacy risk across Codex, Claude Code, or another agent.
Route validation by changed boundaries and risk, choosing the smallest sufficient checks for code, docs, UI, data, auth, sync, build, security, and release work. Use when deciding test commands, reducing unnecessary full-suite runs, auditing test layering, or documenting what was and was not verified.
Repair Codex Desktop sessions that fail after switching model providers or importing/forking old conversations. Use this skill whenever a user mentions an old Codex session becoming invalid, “model provider not found”, “Codex cannot load config.toml”, remote context compaction 404s, “Item with id rs_... not found”, repeated reconnecting during resume, or a migrated Codex conversation that cannot continue—even when the user only asks to inspect or explain the failure. Diagnose the target session across its JSONL rollout, root state_5.sqlite, config.toml, and logs_2.sqlite, then make a target-scoped backup-first repair with verification and restart instructions.
Internationalize the user-facing language layer of a GitHub repository, including its repository description, repository topics, README files, release notes, changelogs, and other explicitly selected documentation, with English as the default and any user-requested locales as mirrors. Use when a user asks to translate, localize, add language mirrors, or check cross-locale parity for a GitHub repository. Confirm the exact document and metadata scope when it is not explicit, identify existing locales, preserve facts, code, commands, links, images, and structure, and provide local Markdown plus GitHub-style previews. Do not use for product UI or runtime i18n, database content, visual README redesign or assets, public-release suitability or security scans, Issue or PR label taxonomies, Git tags, or Git, PR, release, deployment, or production operations.
Review a public repository release across version and tag consistency, final build artifacts, archives, secrets and internal material, third-party licenses, production response headers, and GitHub Release attachments. Use when the user asks to publish, create a tag or Release, verify a launch, or run the complete public-release gate, adapting checks to the project's configuration.
Analyze local Codex session telemetry and produce privacy-safe efficiency reports with actionable workflow recommendations. Use when the user asks to run Codex Doctor, inspect Codex productivity, diagnose context bloat, repository exploration, validation or retry patterns, compare recent sessions, or interpret a codex-efficiency-report HTML/JSON file. Default to local telemetry-only analysis; use the optional compact AI diagnosis only when the user explicitly requests AI interpretation.