Skip to content

dongshuyan/compass-skills / skills/run-history-skill-builder

run-history-skill-builder

Verified

Turn a completed task, browser flow, artifact pipeline, failure-recovery trace, or repeatedly refined workflow into a new reusable skill package or a reviewed skill-design plan. Use when the user asks to make a new skill from real run history, extract a reusable workflow from conversation/logs/files, summarize lessons into a new skill, or produce a plan before writing files. Do not use to upgrade an existing skill or to execute the business workflow itself.

references/skill-design-protocol.md

View on GitHub
# Skill Design Protocol

Read this file after the workflow is reconstructed and before writing the package.

## Choose The Smallest Correct Package

Start with:

```text
<skill-name>/
└── SKILL.md
```

Add directories only when they have a clear job:

- `references/`: long branch-specific guidance, schemas, recovery notes, or source logs;
- `scripts/`: deterministic or repeated checks;
- `evals/`: trigger, regression, or boundary prompts worth preserving;
- `agents/`: optional UI metadata for hosts that use it;
- `assets/`: templates or files that the resulting skill must copy or consume.

Do not add per-skill `README`, installer scripts, or changelogs unless the user or release target explicitly needs them.

## Map Structure Owners

| Problem | Best owner |
| --- | --- |
| Trigger wording | `SKILL.md` frontmatter `description` |
| Main workflow and safety gates | `SKILL.md` body |
| Long branch details | `references/` |
| Deterministic checks | `scripts/` |
| Complex behavioral examples | `references/examples.md` |
| Trigger and regression prompts | `evals/evals.json` |

## Route Design

Keep routes few, strong, and verifiable.

- User-owned decisions: ask the user.
- Machine-checkable facts: script them, test them, or read them from files, diffs, schema, or validators.
- Weak branches such as "it seems okay" or "risk looks low" should be deleted or hardened into explicit gates.

## State Gates

Most packaged workflows need named checkpoints such as:

- input resolved;
- context understood;
- metadata ready;
- validation passed;
- risky action approved;
- success verified.

Do not claim later states without evidence for the earlier ones.

## Examples

Add one-shot or few-shot examples only when they preserve complex behavior that simple rules cannot express cleanly.

Every example should state:

- the invariant;
- the variable fields;
- the failure or ambiguity it prevents;
- the boundary where the example stops applying.

## Eval And Validation Strategy

Before shipping, make sure the package has enough checks to answer:

- When should it trigger?
- What files or resources are required?
- What should block risky steps?
- What proves success?
- What does the final response owe the user?

## Release Discipline

- Strip private paths, credentials, session data, and unrelated internal notes.
- Replace machine-specific examples with placeholders such as `<workspace>` or `<final-artifact>`.
- Keep commands portable: mention `python3`, `python`, or `py -3`, not one fixed runtime only.
- If the user asked for plan-only, stop before file creation and say so explicitly.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall