Skip to content

drafael/coding-harness / skills/clarify

clarify

Verified

Investigate expected-versus-actual behavior, suspected defects, and historical design rationale. Use when someone asks why observed behavior differs from expectations, whether behavior is wrong, or why a design decision exists. Use explain-code for neutral questions about what code does or how it works, and wrong when the user rejects the assistant's current approach.

FileKindSize
LICENSEother1.1 KB
README.mdmarkdown2.0 KB
SKILL.mdskill md6.9 KB
references/evidence-and-confidence.mdmarkdown4.2 KB
references/historical-rationale.mdmarkdown5.0 KB
# Clarify skill

Investigates mismatches between expected and actual behavior, explains current mechanics, and distinguishes intended behavior from configuration, documentation, implementation, and design problems.

It can also investigate historical design rationale when the question asks why a decision, threshold, workaround, or constraint exists. Use `explain-code` for neutral walkthroughs of current mechanics. If the user is rejecting the assistant’s current answer or approach, use `wrong`.

## Example requests

- "I don’t understand why this endpoint returns 404 only in production."
- "Shouldn’t this configuration flag enable retries?"
- "I expected X but got Y. What is happening?"
- "Why was this limit set to 100?"
- "Is this behavior intentional or a regression?"

## Structure

- `SKILL.md` contains the compact routing, diagnosis, explanation, and assessment workflow.
- `references/historical-rationale.md` contains the scalable design-archaeology playbook.
- `references/evidence-and-confidence.md` contains confidence levels, citation discipline, contradiction handling, and gap reporting.

`SKILL.md` is self-sufficient for routine expectation mismatches. Supporting references are intended for on-demand loading when the host supports it. The instructions do not depend on a named model, provider, tool, command syntax, runtime, or agent harness.

## Design principles

- Treat misunderstanding and defects as competing hypotheses.
- Separate current mechanics from historical intent.
- Support claims with evidence and calibrate inferences.
- Preserve contradictions and admit unknowns.
- Scale investigation effort to impact and ambiguity.
- Clarify before changing anything.

## Origins and license

This vendor-neutral compilation adapts ideas from:

- [Umputun’s `clarify`](https://github.com/umputun/cc-thingz/tree/master/plugins/workflow/skills/clarify).
- [Cursor pstack `why`](https://github.com/cursor/plugins/tree/main/pstack/skills/why).

Both upstream works are distributed under the MIT License. See `LICENSE` for the retained notices.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 321kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 197kUpdated Claude CodeCodexVerifiedInstall