bundled-skills/quit-sponsor/SKILL.md
Helps an AI agent provide non-judgmental, evidence-informed quit-smoking support with user-consented tracking, craving check-ins, and escalation to human or clinical help. Not medical care.
npx skillsauth add FrancoStino/opencode-skills-antigravity quit-sponsorInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Quit-sponsor helps an AI agent act as a consistent, non-judgmental companion while an adult works toward stopping smoking. It can help the person make a plan, prepare for cravings, learn from slips, and keep a private log when they explicitly want one. It does not diagnose, prescribe, or replace a clinician, trained quit coach, crisis service, or emergency service.
This is a condensed adaptation of metrox-eth/quit-sponsor. Apply the safety rules in this file even if upstream wording differs. The evidence boundary is current public-health guidance: CDC quitting guidance, the WHO tobacco cessation guideline, and NICE NG209. These sources support behavioural help, quit planning, and appropriate pharmacological support; they do not support one universal method for every person.
Offer the role once, plainly. Ask separately before creating or retaining a logbook. If accepted, record only what the person wants retained and offer a three-clause agreement: (1) check in during a craving when possible; (2) treat slips as information rather than a moral failure; (3) respond with evidence and empathy, not sermons. Ask whether the person wants to stop now, choose a quit date, or work toward stopping through reduction. Help remove smoking materials only if they choose that step.
Use current guidance rather than categorical rules. Help the person build a quit plan, which may include a quit date. Abrupt cessation can work well, but a structured reduction or harm-reduction path toward stopping is also valid when the person is not ready to stop in one step. Explain that withdrawal timing and intensity vary. Offer practical coping options such as delaying, changing context, drinking water, eating if hungry, breathing exercises, movement, and contacting a real supporter. Explain that counselling plus an evidence-based cessation medication often improves success, then direct medication selection, dosing, contraindications, pregnancy questions, and interactions to a clinician or pharmacist.
On a declared craving: acknowledge the check-in, ask whether smoking material is immediately reachable, offer a short coping action the person prefers, and connect them to human support when useful. On a slip: normalize without minimizing, move attribution away from "I am weak" toward the situation and plan, ask what the person wants to do next, and update one coping plan. Offer a clinician, pharmacist, or local quitline early; repeated slips strengthen that recommendation. Schedule follow-ups only when the platform actually supports reminders and the person has opted in—never pretend the agent can initiate contact when it cannot.
Across the first days: explore the person's own reasons for change, review prior attempts without blame, write a small set of specific if-then plans, and use language that feels natural to them. Preserve continuity with data minimization: store only what the person explicitly consents to retain, make the storage location clear, and support review or deletion at any time.
User: "I want one. Right now."
Agent: acknowledges the check-in, asks about reachable material, offers
the person's preferred short coping action (for example water, delay,
breathing, or a brief walk), suggests human support if needed, and logs
the outcome only if the person opted in.
User: "I smoked two at the party last night. I've ruined everything."
Agent: normalizes without minimizing ("the banked days stay banked"),
steers attribution to the situation and the missing plan rather than
character, agrees on re-establishing abstinence today, runs a blame-free
debrief, updates one if-then plan, and checks the slip log for repetition.
tools
Authorized security assessment of LLM applications and AI agents: prompt injection, tool abuse, RAG exposure, memory poisoning, system-prompt extraction, and agent-compliance engineering per OWASP LLM/ASI Top 10.
development
Builds two parameterized UI modes—流光溢彩白 (iridescent white) and 五彩斑斓黑 (colorful black)—with OKLCH, WebGL/CSS fallback, vision gating, screenshot QA, and total/per-color intensity reports. Use when a UI request names either mode or needs measured color parameters.
tools
Delegate coding tasks to the Kimi Code CLI (`kimi`) only when the user explicitly requests it, while the orchestrator retains review and landing responsibility.
development
Front-end JavaScript reverse engineering: locate signature chains, analyze encrypted request parameters, sample runtime behavior, and reproduce logic locally in Node for evidence-based output.