ysskrishna/ai-agent-skills / skills/ethical-thinking
ethical-thinking
VerifiedUse for ethical thinking, or when a plan could harm or unfairly burden people: "should we do this?", "is this fair?", "privacy or bias risk", "who gets hurt?". Maps stakeholders, harms, power and consent. Skip for legal advice, neutral fact lookups and implementation-only work.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.1 KB |
| references/example.md | markdown | 2.1 KB |
references/example.md
View on GitHub# Example: Ethical Thinking (light path)
Illustrative case. All details are invented.
**User:** "We want to use customers' support chat transcripts to train an internal model that drafts replies. Is that okay?"
**Focal action:** train an internal reply-drafting model on past customer support transcripts.
**Pass:** Stakeholders, Values, Harms and Benefits, Justice and Power, Options, Recommendation.
**Stakeholders**
- Direct: customers whose chats are used; support agents whose replies become training data.
- Indirect: future customers who receive model-drafted replies.
- Vulnerable: customers who shared health or financial details in chat, because they wrote under the assumption that a person would read it.
**Values**
- **Autonomy vs efficiency:** customers did not choose this use; faster replies help everyone.
- **Privacy vs quality:** more complete transcripts give a better model but carry more personal data.
**Harms and Benefits**
- Harm: the model repeats personal details in another customer's reply. Plausible if transcripts are not scrubbed. Hard to undo once trained.
- Benefit: faster and more consistent replies. Plausible; benefits depend on review by agents.
- Predicted only; no incident history was provided.
**Justice and Power**
- Customers carry the privacy risk; the company gets the benefit. Customers cannot say no after the fact.
- Agents' work becomes training data without a stated say. Check what their contract and policy allow.
**Options**
> **Option:** do not proceed - **Value fit:** respects autonomy - **Residual harm:** slower replies - **Safeguards:** none needed.
> **Option:** proceed with redaction and opt-out - **Value fit:** balances both - **Residual harm:** redaction misses - **Safeguards:** automated plus sampled manual checks, opt-out honored, retention limit, owner named.
**Recommendation**
- Conditional: proceed only with redaction, an opt-out and a named owner. Legal review needed on consent wording and data rules.
- Strongest reason against: transcripts may hold sensitive details that redaction cannot reliably find.
- Monitor: leaks found in drafts, opt-out rate, and agent feedback.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |