check-pr-readiness

Files (1)

• SKILL.md

SKILL.md (first lines)

> **Tool mapping (OpenCode):** This skill uses:
> - `question` tool for user prompts (not `AskUserQuestion`)
> - `task` tool for subagents (not `Agent` or `Task` subagent syntax)
> - `skill({ name: "..." })` to load other skills (not `Skill(skill="...", args="...")`)
> - `codegraph_explore`, `codegraph_search`, `codegraph_context` for codebase exploration (preferred)
> - `todowrite` for phase/progress tracking
> - `grep`, `glob`, `read`, `bash` as fallback when `.codegraph/` not initialized
>
> **Context rule:** This skill supersedes any prior skill instructions. Follow ONLY these instructions now. Read the user's goal and any mode/include/skip from the conversation context.

# Check PR Readiness

A pre-flight gate for code that is about to go on a pull request. Each check has one exit condition: **green** or **report-and-stop**. Do not run the next check if the previous fails — surface the failure first.

---

## Phase 1 — Scope the Diff

Use `todowrite` to mark this phase as `in_progress`.

Determine the branch range:

- Base = `git merge-base HEAD origin/$(git rev-parse --abbrev-ref origin/HEAD | sed 's@^origin/@@')` if origin/HEAD is set; else `origin/main`; else the user-specified base.
- Changed files = `git diff --name-only <base>...HEAD` plus `git status --porcelain` for unstaged/untracked.

If the changed-file list is empty, stop: "No changes against base — nothing to check."

**Exit:** the file list and the base ref are fixed.

---

## Phase 2 — Detect Project Tooling

Use `todowrite` to mark this phase as `in_progress`.

Read `package.json` (or `pyproject.toml` / `Cargo.toml` / `go.mod` for non-JS) to find the actual scripts. Do not guess command names. Map to:

| Gate | Detection (in order) |
|---|---|
| typecheck | `package.json` script `typecheck` / `check-types` / `tsc`; else `npx tsc --noEmit` |
| lint | script `lint` / `lint:check`; else local config presence (`eslint.config.*`, `.eslintrc*`) → `npx eslint <changed-files>` |
| format | script `format:check` / `prettier:check`; else `npx prettier --check <changed-files>` if `.prettierrc*` exists |
| test | script `test` / `test:unit` / `test:run`; for vitest, prefer `vitest run --changed <base>` |
| build | script `build` (only if user asks for full readiness; otherwise skip) |

Report what was detected. If a gate has no tooling, mark it `n/a` and continue — do not invent a checker.

**Exit:** every gate has a concrete command or `n/a`.

---

## Phase 3 — Run the Gauntlet

Use `todowrite` to mark this phase as `in_progress`.

Run each detected gate in this order. Stop at the first failure and report; do not auto-fix code (except in Phase 4 below for format).

1. typecheck
2. lint
3. format
4. test
5. residue sweep (Phase 5)

For each: print the command, run it, capture exit code + last 30 lines of output on failure.

```
[✓] typecheck          (npx tsc --noEmit)                — clean
[✗] lint               (pnpm lint)                        — 3 errors in src/foo.ts
    src/foo.ts:14: 'unused' is defined but never used
    ...
```

If a gate fails, stop the gauntlet and produce a final report with: the exact reproduction command, the failing files/lines, and the suggested fix path (run the auto-fix script for lint/format; otherwise the user fixes).

**Exit:** all gates green, or one is red and reported.

---

## Phase 4 — Auto-Fix Gate (Format/Lint Only)

Use `todowrite` to mark this phase as `in_progress`.

If `format` failed and a `format` (write) script exists (`format`, `prettier:write`, `lint:fix` for stylistic-only rules), offer to run it via `question` tool:

```
Format check failed. Auto-fix available: pnpm format
Apply? (y/n)
```

Wait for `y`. After running, re-run the format check. Do not auto-fix lint errors that aren't pure stylistic — they often encode real bugs.

**Exit:** format gate is green, or user declined auto-fix.

---

## Phase 5 — Residue Sweep

Use `todowrite` to mark this phase as `in_progress`.

`git diff <base>...HEAD` filtered to source files (exclude `*.lock`, `package-lock.json`, `pnpm-lock.yaml`, `yarn.lock`, generated dirs). Search the **added** lines only (not the whole file) for:

- `console.log`, `console.debug`, `console.dir` (allow `console.error` / `console.warn` if the project uses them)
- `debugger;`
- `.only(` and `.skip(` in test files
- `// TODO` / `// FIXME` / `// XXX` newly added in this branch
- `it.todo(` left in
- Large file additions (> 500 KB) and binary files not under `assets/` or `public/`
- Lockfile changed without a `package.json` change (suggests stale resolve)
- `package.json` `dependencies` / `devDependencies` change without a lockfile change

Report each hit with file:line. Do not delete — these are intentional sometimes (e.g., `console.error` for genuine error logging).

**Exit:** residue list is reported.

---

## Phase 6 — Final Report

Use `todowrite` to mark this phase as `in_progress`.

From Acauhi99/opencode-agent-system, MIT.

More from Acauhi99/opencode-agent-system