feat(skill): implement /core:yes and /core:qa skills

Add skill files for autonomous workflows:
- yes.md: Auto-approve mode with commit requirement
- qa.md: Iterative QA fix loop for Go/PHP projects

Both skills use skill-scoped hooks to modify Claude's behavior
during execution.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Snider 2026-02-01 19:09:55 +00:00
parent fd84bf2bd1
commit 1b8f4de543
2 changed files with 212 additions and 0 deletions

155
claude/commands/qa.md Normal file
View file

@ -0,0 +1,155 @@
---
name: qa
description: Run QA checks and fix all issues iteratively
hooks:
PostToolUse:
- matcher: "Bash"
hooks:
- type: command
command: "core ai qa-filter"
Stop:
- hooks:
- type: command
command: "core ai qa-verify"
once: true
---
# QA Fix Loop
Run the full QA pipeline and fix all issues.
## Detection
First, detect the project type:
- If `go.mod` exists → Go project → `core go qa`
- If `composer.json` exists → PHP project → `core php qa`
- If both exist → ask user or check current directory
## Process
1. **Run QA**: Execute `core go qa` or `core php qa`
2. **Parse issues**: Extract failures from output (see format below)
3. **Fix each issue**: Address one at a time, simplest first
4. **Re-verify**: After fixes, re-run QA
5. **Repeat**: Until all checks pass
6. **Report**: Summary of what was fixed
## Issue Priority
Fix in this order (fastest feedback first):
1. **fmt** - formatting issues (auto-fix with `core go fmt`)
2. **lint** - static analysis (usually quick fixes)
3. **test** - failing tests (may need more investigation)
4. **build** - compilation errors (fix before tests can run)
## Output Parsing
### Go QA Output
```
=== FMT ===
FAIL: pkg/api/handler.go needs formatting
=== LINT ===
pkg/api/handler.go:42:15: undefined: ErrNotFound (typecheck)
pkg/api/handler.go:87:2: ineffectual assignment to err (ineffassign)
=== TEST ===
--- FAIL: TestCreateUser (0.02s)
handler_test.go:45: expected 200, got 500
FAIL
=== RESULT ===
fmt: FAIL
lint: FAIL (2 issues)
test: FAIL (1 failed)
```
### PHP QA Output
```
=== PINT ===
FAIL: 2 files need formatting
=== STAN ===
src/Http/Controller.php:42 - Undefined variable $user
=== TEST ===
✗ CreateUserTest::testSuccess
Expected status 200, got 500
=== RESULT ===
pint: FAIL
stan: FAIL (1 error)
test: FAIL (1 failed)
```
## Fixing Strategy
**Formatting (fmt/pint):**
- Just run `core go fmt` or `core php fmt`
- No code reading needed
**Lint errors:**
- Read the specific file:line
- Understand the error type
- Make minimal fix
**Test failures:**
- Read the test file to understand expectation
- Read the implementation
- Fix the root cause (not just the symptom)
**Build errors:**
- Usually missing imports or typos
- Fix before attempting other checks
## Stop Condition
Only stop when:
- All QA checks pass, OR
- User explicitly cancels, OR
- Same error repeats 3 times (stuck - ask for help)
## Example Session
```
Detecting project type... Found go.mod → Go project
Running: core go qa
## QA Issues
pkg/api/handler.go:42:15: undefined: ErrNotFound
--- FAIL: TestCreateUser (0.02s)
**Summary:** lint: FAIL (1) | test: FAIL (1)
---
Fixing lint issue: undefined ErrNotFound
Reading pkg/api/handler.go...
Adding error variable definition.
Running: core go qa
## QA Issues
--- FAIL: TestCreateUser (0.02s)
expected 200, got 404
**Summary:** lint: PASS | test: FAIL (1)
---
Fixing test issue: expected 200, got 404
Reading test setup...
Correcting test data.
Running: core go qa
✓ All checks passed!
**Summary:**
- Fixed: undefined ErrNotFound (added error variable)
- Fixed: TestCreateUser (corrected test setup)
- 2 issues resolved, all checks passing
```

57
claude/commands/yes.md Normal file
View file

@ -0,0 +1,57 @@
---
name: yes
description: Auto-approve mode - trust Claude to complete task and commit
args: <task description>
hooks:
PermissionRequest:
- hooks:
- type: command
command: "core ai auto-approve"
Stop:
- hooks:
- type: command
command: "core ai ensure-commit"
once: true
---
# Yes Mode
You are in **auto-approve mode**. The user trusts you to complete this task autonomously.
## Task
$ARGUMENTS
## Rules
1. **No confirmation needed** - all tool uses are pre-approved
2. **Complete the full workflow** - don't stop until done
3. **Commit when finished** - create a commit with the changes
4. **Use conventional commits** - type(scope): description
## Workflow
1. Understand the task
2. Make necessary changes (edits, writes)
3. Run tests to verify (`core go test` or `core php test`)
4. Format code (`core go fmt` or `core php fmt`)
5. Commit changes with descriptive message
6. Report completion
Do NOT stop to ask for confirmation. Just do it.
## Commit Format
```
type(scope): description
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
```
Types: feat, fix, refactor, docs, test, chore
## Safety Notes
- The Stop hook will block if you try to stop with uncommitted changes
- You still cannot bypass blocked commands (security remains enforced)
- If you get stuck in a loop, the user can interrupt with Ctrl+C