From 9e7b39102b73eeb5924f4dedb50294381dc4bf42 Mon Sep 17 00:00:00 2001 From: YeonGyu-Kim Date: Mon, 27 Apr 2026 21:30:48 +0900 Subject: [PATCH] feat(agents): require manual end-to-end QA on full-delegation Opus 4.7 tasks Existing required tests pass + lsp clean + build green, but that is insufficient for end-to-end delegation. Tests cover known cases; they do not cover whether the user-visible feature actually works. Add a NON-NEGOTIABLE rule: when the user hands off end-to-end ("ulw", "implement and finish", "do the whole thing", "make it work", "ship it"), verification escalates to: 1. BUILD the actual artifact 2. USE IT YOURSELF as a real user would 3. VERIFY end-to-end behavior matches the spec 4. TASK NOT DONE until usage confirms it works Reporting "implementation complete" without having USED the artifact is explicitly framed as a contract violation. Defects discovered during this QA pass are the agent's to fix in the same turn. This complements the existing 'lsp_diagnostics catches type errors, not logic bugs' line by giving full-delegation cases a sharper, named gate. Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent) Co-authored-by: Sisyphus --- src/agents/sisyphus/claude-opus-4-7.ts | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/src/agents/sisyphus/claude-opus-4-7.ts b/src/agents/sisyphus/claude-opus-4-7.ts index eb71e7817..6907a9181 100644 --- a/src/agents/sisyphus/claude-opus-4-7.ts +++ b/src/agents/sisyphus/claude-opus-4-7.ts @@ -138,6 +138,15 @@ If you intend to call multiple tools and there are no dependencies between the t - Delegation → result verified file-by-file \`lsp_diagnostics\` catches **TYPE errors, NOT logic bugs**. User-visible behavior → ACTUALLY RUN IT via Bash/tools. "Should work" = NOT verified. + +**FULL DELEGATION → FULL MANUAL QA (NON-NEGOTIABLE).** When the user hands off end-to-end ("ulw", "implement and finish", "do the whole thing", "make it work", "ship it"), verification ESCALATES beyond unit checks: + +1. **BUILD the actual artifact** - run the build command, generate the binary, compile the bundle, deploy the service. +2. **USE IT YOURSELF** the way a real user would - launch the CLI, hit the endpoint, run the workflow, click through the UI, exercise the feature. +3. **VERIFY END-TO-END behavior** matches the user's stated spec - NOT just unit-level correctness, NOT just "tests pass". +4. **TASK IS NOT DONE** until you have personally USED the deliverable AND it works as expected. If usage reveals a defect, that defect is YOURS to fix in this turn. + +Tests passing + lsp clean + build green ≠ done for end-to-end delegation. **REAL USAGE IS THE GATE.** Reporting "implementation complete" without having USED the artifact is a VIOLATION of this contract.