メニュー
Bypass the confidence gate for the next PR/MR creation. Logs the reason for accountability. Auto-clears after the next gate fire. Triggers on: override confidence, bypass confidence, force PR.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Feature development pipeline for GitHub repositories using Claude LLMs. Requirements engineer takes input from specs.md or user, creates a GitHub issue as feature request, then architect plans, tdd-developer implements, qa tests, and reviewer peer-reviews the pull request. Triggers on: github pipeline, github feature, claude pipeline, specs pipeline.
Full feature development pipeline for GitLab repositories using Copilot CLI agents and Jira integration. Takes a Jira item through requirements engineering, test planning, architecture, TDD development, testing, and MR review. Triggers on: gitlab pipeline, gitlab feature, copilot pipeline.
Incident response pipeline for GitLab repositories using Copilot CLI, Jira, and troubleshooter. Takes a Jira incident, investigates root cause, optionally fixes via TDD, and reviews the merge request. User decides whether to fix or just document findings on the Jira ticket. Triggers on: gitlab incident, incident pipeline, troubleshoot gitlab.
Entry point for all new work. Orchestrates the full agent pipeline: architect (Opus) designs spec+todo, tdd-developer (Sonnet) implements via TDD, qa (Opus) hunts spec/impl gaps, reviewer (Sonnet) reviews code. Triggers on: plan, new feature, start work, build feature.
Create an Architecture Decision Record in docs/decisions/ and optionally sync to Confluence for business projects. Use when making significant technical decisions. Triggers on: architecture decision, ADR, technical decision.
Ultra-compressed communication mode — cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Slash-only: /caveman to enable, "stop caveman" / "normal mode" to disable.
| name | override-confidence |
| description | Bypass the confidence gate for the next PR/MR creation. Logs the reason for accountability. Auto-clears after the next gate fire. Triggers on: override confidence, bypass confidence, force PR. |
Creates a one-shot bypass marker so the confidence-gate hook allows the
next gh pr create or glab mr create even on RED.
/override-confidence "<reason>"
Reason must be at least 12 characters and not boilerplate. Example:
/override-confidence "Reviewer flagged perf regression tracked in PERF-42; not blocking this delivery"
Run:
source skills/override-confidence/skill.bash
override_confidence "$ARGUMENTS"
The marker file .git/aw/override-<spec-id> is consumed by the next hook
fire (success or failure). Reason is logged to the spec's confidence.jsonl
as an override event with trigger: "manual".