Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user asks for help with Claude Code, the Claude Code CLI, Anthropic coding-agent workflows, prompts to hand off work to Claude Code, or troubleshooting Claude Code setup and terminal usage.
.claude/skills/hashgraph-online-claude-code/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 137% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 18% | 0% |
Use this skill when the user wants Codex to help them work with Claude Code.
claude is available with claude --version.claude_code_status first.git status --short when operating inside a repo.README.md, CLAUDE.md, AGENTS.md, or package scripts before suggesting commands.This plugin never stores Claude credentials. It accesses a Claude Code account only through the local claude CLI after the user has authenticated with Anthropic.
Use this setup sequence:
claude_code_status or claude --version.irm https://claude.ai/install.ps1 | iexnpm install -g @anthropic-ai/claude-codeclaude_code_login to open the official browser login and complete the confirmation.claude_code_status.claude_code_prompt only after the CLI is installed and authenticated.claude_code_prompt runs only in the current Codex task workspace, always starts Claude Code with --safe-mode, and defaults to manual permissions. This prevents repository configuration, hooks, plugins, and MCP configuration from being loaded implicitly.
Use manual when each action should ask for approval, or plan when the task must remain read-only. Autonomous modes, including dontAsk, auto, and acceptEdits, are intentionally unavailable through this plugin. Do not use bypass-permissions modes from this plugin.
To switch accounts or refresh auth, call claude_code_login, or run claude and use /login or /logout in the Claude Code session.
Use claude_code_prompt with permissionMode: "plan". Include the review target, relevant files, constraints, and expected verification. Ask Claude Code to report findings before suggestions.
Use plan when the user wants analysis without edits. Use manual only when the user explicitly wants an interactive implementation where every action still requires approval.
Run claude_code_status, use claude_code_login only when authentication needs attention, then run status again. Keep installation problems separate from project-specific failures.
textGoal: <one concrete outcome> Context: - Repo: <path or project name> - Relevant files: <files/directories> - Constraints: <user constraints> Task: <specific implementation or investigation request> Verification: - <command or manual check> - <command or manual check>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 18,963 | 12,947 | -32% | 1 | 1 | 0% | 2,104 | 2,108 | +0% | 0 | 0 | — |
case-01 | fail→fail | 16,703 | 20,215 | +21% | 1 | 1 | 0% | 327 | 1,397 | +327% | 0 | 0 | — |
case-02 | fail→fail | 14,797 | 14,143 | -4% | 1 | 1 | 0% | 1,661 | 1,251 | -25% | 0 | 0 | — |
case-03 | pass→fail | 15,042 | 12,519 | -17% | 1 | 1 | 0% | 1,289 | 1,170 | -9% | 0 | 0 | — |
case-04 | fail→pass | 15,714 | 10,920 | -31% | 1 | 1 | 0% | 1,745 | 1,824 | +5% | 0 | 0 | — |
case-06 | fail→fail | 14,658 | 14,500 | -1% | 1 | 1 | 0% | 2,013 | 2,276 | +13% | 0 | 0 | — |
case-07 | fail→pass | 10,320 | 7,088 | -31% | 1 | 1 | 0% | 723 | 1,246 | +72% | 0 | 0 | — |
case-08 | pass→pass | 6,176 | 8,342 | +35% | 1 | 1 | 0% | 838 | 1,479 | +76% | 0 | 0 | — |
case-09 | fail→pass | 25,852 | 2,958 | -89% | 1 | 1 | 0% | 1,777 | 1,359 | -24% | 0 | 0 | — |
case-10 | fail→pass | 3,550 | 3,128 | -12% | 1 | 1 | 0% | 560 | 1,327 | +137% | 0 | 0 | — |
case-11 | fail→pass | 15,567 | 10,149 | -35% | 1 | 1 | 0% | 1,475 | 1,735 | +18% | 0 | 0 | — |
case-12 | pass→pass | 12,675 | 6,176 | -51% | 1 | 1 | 0% | 1,876 | 1,742 | -7% | 0 | 0 | — |
case-13 | pass→pass | 10,808 | 7,611 | -30% | 1 | 1 | 0% | 1,880 | 1,209 | -36% | 0 | 0 | — |
case-14 | fail→pass | 8,768 | 6,939 | -21% | 1 | 1 | 0% | 984 | 1,191 | +21% | 0 | 0 | — |
case-15 | fail→pass | 15,806 | 3,731 | -76% | 1 | 1 | 0% | 1,647 | 1,482 | -10% | 0 | 0 | — |
case-16 | fail→pass | 10,666 | 6,194 | -42% | 1 | 1 | 0% | 1,621 | 1,968 | +21% | 0 | 0 | — |
case-17 | pass→pass | 17,820 | 6,177 | -65% | 1 | 1 | 0% | 1,844 | 1,729 | -6% | 0 | 0 | — |
case-18 | pass→pass | 12,632 | 8,673 | -31% | 1 | 1 | 0% | 1,121 | 1,510 | +35% | 0 | 0 | — |
case-19 | pass→pass | 14,816 | 7,811 | -47% | 1 | 1 | 0% | 1,444 | 1,299 | -10% | 0 | 0 | — |
case-20 | fail→pass | 9,237 | 6,548 | -29% | 1 | 1 | 0% | 1,540 | 1,100 | -29% | 0 | 0 | — |
case-21 | pass→pass | 13,541 | 10,216 | -25% | 1 | 1 | 0% | 1,674 | 1,900 | +14% | 0 | 0 | — |
case-22 | pass→pass | 14,035 | 12,225 | -13% | 1 | 1 | 0% | 1,656 | 2,368 | +43% | 0 | 0 | — |
case-23 | pass→pass | 19,346 | 14,497 | -25% | 1 | 1 | 0% | 2,226 | 2,966 | +33% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 20 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +35 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.