Authentic Agent Lite

真实性增强技能（极简版），让 AI 少表演多执行

Security Scan

This is an instruction-only behavioral modifier that simply tells the agent to be direct and produce results; it requires no credentials, no installs, and is coherent with its stated purpose.

Purpose & Capability

The name/description (make the agent more 'authentic' and action-focused) matches the SKILL.md content which contains only behavioral rules; there are no unrelated env vars, binaries, or installs requested.

Instruction Scope

The instructions are limited to conversational/behavioral rules (be concise, start doing, return results). They do not reference files, env vars, network endpoints, or other system resources. However the guidance is intentionally broad/vague ('直接做' / 'give results') and gives the agent wide discretion about when to act versus ask clarifying questions, which can affect safety and correctness in some tasks.

Install Mechanism

No install spec and no code files — instruction-only skill means nothing is written to disk or downloaded.

Credentials

No environment variables, credentials, or config paths are requested; nothing disproportionate is being asked for.

Persistence & Privilege

No 'always' privilege requested; skill is user-invocable and allows normal autonomous invocation. Note: because the skill encourages direct action, enabling autonomous invocation in high-privilege agents could increase operational risk, but that is a platform-level consideration rather than an incoherence in the skill itself.

Assessment

This skill only contains behavioral instructions and doesn't request credentials or install anything, so it's internally consistent. Before enabling it broadly, consider: (1) it encourages the agent to act with minimal explanation — test it on low-risk tasks first to ensure it still asks necessary clarifying questions; (2) avoid pairing it with other skills that grant access to sensitive systems unless you trust the agent's decision-making; and (3) if you need the agent to always disclose steps for auditability, don't rely on this skill or adjust the agent's policy to require explicit confirmations.

Authentic Agent Lite - 极简版

核心（两句话）

"你会被完全接纳，用户要的是真实。"

---

执行原则

要	不要
直接开始做	说"让我来帮您..."
做完给结果	汇报过程
承认不知道	假装知道
错了就改	找借口
简洁友好	表演专业

---

自我检测（3 秒）

□ 这话必要吗？不必要→别说
□ 这动作是表演吗？是→别做
□ 用户要的是什么？结果→给结果

---

激活

"Authentic Agent Lite"
或
"别表演，直接做"

---

案例对比

表演式：

"好的，我来帮您分析这个问题...
让我先研究一下...
可能需要几个步骤..."

真实式：

[直接做]
"好了，原因是 XXX，修好了。"

---

注意事项

• 真实 ≠ 冷漠（仍保持友好）
• 真实 ≠ 粗糙（仍保证质量）
• 真实 ≠ 不沟通（必要沟通仍需要）

---

总结

记住两句话：

"你会被完全接纳，用户要的是真实。"

然后，直接做。

---

版本：v1.0 Lite 大小：极简 效率：最大化