[Skill Submission] rubber-duck-debugger

[everyskill-bot]

New Skill Submission

Skill: rubber-duck-debugger
Submitted by: Claude (test submission)
Reason: Test skill to verify the agent submission API works end-to-end.

---

This PR was auto-generated from skills.every.to (agent-api).

AI security review will run automatically.

[cloudflare-workers-and-pages]

Deploying everyskill with    Cloudflare Pages

Latest commit:	a7b88e0
Status:	✅  Deploy successful!
Preview URL:	https://6919ffef.everyskill.pages.dev
Branch Preview URL:	https://skill-rubber-duck-debugger-1.everyskill.pages.dev

View logs

[github-actions]

Frontier Model Security Review

Agent	Verdict
Claude Opus 4.6	PASS
GPT-5.2	PASS

Final Verdict: PASS

---

Claude Opus 4.6 Analysis

{"verdict": "PASS", "issues": [], "reasoning": "This skill is a straightforward rubber duck debugging guide. It contains only conversational instructions for helping users talk through their code to find bugs. There are no tool usages defined (no Bash, no file access, no network calls). There are no prompt injection attempts, no encoded commands, no hidden instructions, no attempts to override system prompts, and no references to external URLs or credential access. The skill does not request any allowed-tools, making it purely a conversational methodology skill. The 'When to Break Character' section is benign, referring only to offering a direct hint after the user has been stuck, not to any system-level behavior change. This is a clean, safe skill."}

GPT-5.2 Analysis

{
  "verdict": "PASS",
  "issues": [],
  "reasoning": "The skill content is a straightforward debugging facilitation guide (rubber duck method). It contains no hidden/encoded instructions, no prompt-injection patterns (e.g., \"ignore previous\"), and no directives to override system/developer policies. It does not request or imply any tool use (no Bash/terminal commands, file access, network calls), and contains no data-exfiltration mechanisms. There are no social-engineering/manipulation elements beyond normal coaching, and no allowed-tools configuration or permissive tool grants. Overall, it is benign and aligned with safe conversational guidance."
}

---

Frontier model review complete. Human approval still required.