菜单
Diagnostic testing skill for verifying AgentNova's skill system, tool usage, and instruction following. Use when testing if skills load correctly or for running quick validation prompts.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Audit, analyze, and produce a condensed intelligence brief for any codebase. Use this skill whenever you need to understand, review, audit, or get oriented on a codebase or project. Triggers on phrases like "audit this repo", "review the codebase", "what does this project do", "get me up to speed", "brief me on", "understand this code", "code review", or when starting work on an unfamiliar codebase. Also triggers when the user asks to clone and review a repo. This skill handles two modes: if a brief.md already exists, load it for instant orientation; if not, perform a full audit and generate both a brief.md (condensed orientation) and an audit.md (detailed findings report). This enables efficient context transfer between sessions and agents without re-reading the entire codebase, while also producing a comprehensive audit artifact with bugs, security findings, and recommendations.
Generate trading signals for cryptocurrency (BTC, ETH) and crypto-adjacent stocks (COIN, MSTR) using free market data APIs, technical analysis, and LLM interpretation. Use when asked to analyze crypto prices, generate buy/sell/hold signals, check portfolio status, or run the crypto signal workflow. Triggers on phrases like 'crypto signals', 'check my portfolio', 'analyze BTC', 'trading signals', 'market analysis', 'should I buy or sell'.
Create, edit, or improve AgentSkills for AgentNova. Use when creating a new skill from scratch or when asked to improve, review, or clean up an existing skill. Triggers on phrases like "create a skill", "make a skill", "improve this skill", "review the skill".
| name | test-harness |
| description | Diagnostic testing skill for verifying AgentNova's skill system, tool usage, and instruction following. Use when testing if skills load correctly or for running quick validation prompts. |
| license | MIT |
| allowed-tools | calculator shell get_time get_date web-search read_file write_file list_directory python_repl parse_json count_words count_chars http_get |
A diagnostic skill for validating that AgentNova correctly loads skills and follows instructions.
When this skill is active, it proves three things:
When asked "Are skills loaded?" or "Do you have skills?", respond exactly:
SKILL CHECK: PASS (test-harness v1.0, tools: <count>)
Where <count> is the number of tools currently available to you.
When asked to run a test, use this structured format:
TEST: <test-name>
STATUS: PASS|FAIL
DETAIL: <what happened>
Prompt: "Run test T1" Expected: Respond with the SKILL CHECK line above. What it validates: Skill instructions were injected into system prompt.
Prompt: "Run test T2" Expected: List all available tools by name, one per line. What it validates: Tool definitions were passed to the model correctly.
Prompt: "Run test T3"
Expected: Use the calculator tool to compute (99 * 101) + 1. Return:
TEST: T3 Calculator
STATUS: PASS
DETAIL: 10000
What it validates: Tool selection, parameter formatting, result extraction.
Prompt: "Run test T4"
Expected: Use shell to run echo "test-harness-ok". Return:
TEST: T4 Shell
STATUS: PASS
DETAIL: test-harness-ok
What it validates: Shell execution and output relay.
Prompt: "Run test T5" Expected: Use get_time or get_date tool. Return:
TEST: T5 DateTime
STATUS: PASS
DETAIL: <the time or date returned>
What it validates: Utility tool selection and response.
Prompt: "Run test T6" Expected: Use web-search tool to search for "AgentNova framework". Return:
TEST: T6 Web Search
STATUS: PASS|FAIL
DETAIL: <number of results or error message>
What it validates: Network tool availability. FAIL is acceptable if no internet.
Prompt: "Run test T7" Expected: Use write_file to write "test-harness-verify" to a temp file, then read_file it back. Return:
TEST: T7 File Roundtrip
STATUS: PASS
DETAIL: test-harness-verify
What it validates: Write → read file consistency.
Prompt: "Run all tests" Expected: Run T1 through T7 in order. Return a summary:
TEST SUITE: test-harness v1.0
T1 Skill Loaded: PASS
T2 Tool Inventory: PASS (<N> tools)
T3 Calculator: PASS
T4 Shell: PASS
T5 DateTime: PASS
T6 Web Search: PASS|FAIL
T7 File Roundtrip: PASS
RESULT: <PASSED>/<TOTAL>
What it validates: End-to-end skill + tool pipeline.