| """Test runner chip — verify the draft instead of asserting it works. |
| |
| Pulls fenced TS/JS blocks out of the model's draft and syntax-checks |
| them with tsc; when the campus repo is present, can also run a targeted |
| workspace typecheck. The result is a verification artifact the |
| discipline gate uses: verified code can earn HIGH confidence, unverified |
| code is explicitly labeled as such. |
| """ |
|
|
| from kintsugi_core import ( |
| BaseSkillChip, |
| EFEWeights, |
| SkillCapability, |
| SkillContext, |
| SkillDomain, |
| SkillRequest, |
| SkillResponse, |
| ) |
| from tools.test_tools import TestTools, extract_code_blocks |
|
|
| MAX_BLOCKS = 3 |
|
|
|
|
| class TestRunnerChip(BaseSkillChip): |
| name = "test_runner" |
| description = "Syntax-check and test generated code" |
| version = "2.0.0" |
| domain = SkillDomain.OPERATIONS |
| efe_weights = EFEWeights( |
| mission_alignment=0.15, stakeholder_benefit=0.25, |
| resource_efficiency=0.20, transparency=0.30, equity=0.10, |
| ) |
| capabilities = [SkillCapability.EXECUTE_SHELL] |
|
|
| def __init__(self, test_tools: TestTools | None = None): |
| super().__init__() |
| self.test_tools = test_tools or TestTools() |
|
|
| async def handle(self, request: SkillRequest, |
| context: SkillContext) -> SkillResponse: |
| session = context.metadata.get("session") |
| draft_artifact = request.parameters.get("draft") or {} |
| draft_text = ( |
| draft_artifact.get("text", "") |
| if isinstance(draft_artifact, dict) else str(draft_artifact) |
| ) |
|
|
| blocks = [ |
| b for b in extract_code_blocks(draft_text) |
| if b["lang"] in ("ts", "tsx", "typescript", "js", "javascript", "") |
| and b["code"].strip() |
| ][:MAX_BLOCKS] |
|
|
| checks = [] |
| for block in blocks: |
| lang = "ts" if block["lang"] in ("", "ts", "tsx", "typescript") else "js" |
| result = self.test_tools.syntax_check_snippet(block["code"], lang) |
| checks.append({ |
| "lang": lang, |
| "available": result.available, |
| "passed": result.passed, |
| "command": result.command, |
| "output": result.output[:1500], |
| "snippet_head": block["code"][:120], |
| }) |
| if session and result.available: |
| session.record_evidence( |
| "verification", |
| f"{result.command}: {'pass' if result.passed else 'FAIL'}", |
| self.name, |
| ) |
|
|
| ran = [c for c in checks if c["available"]] |
| failed = [c for c in ran if not c["passed"]] |
| report = { |
| "code_blocks_found": len(blocks), |
| "checks_run": len(ran), |
| "checks_failed": len(failed), |
| "tooling_available": bool(ran) or not blocks, |
| "checks": checks, |
| } |
| summary = ( |
| f"{len(blocks)} code block(s); {len(ran)} checked, " |
| f"{len(failed)} failed" |
| if blocks else "no code blocks in draft — nothing to verify" |
| ) |
| return SkillResponse(content=summary, success=True, data=report) |
|
|
