git clone --depth 1 https://github.com/Prism-Shadow/penguin-harness /tmp/0.2.4 && cp -r /tmp/0.2.4/changelog/0.2.4/2026-08-21-agent-initialization- ~/.claude/skills/0.2.42026-08-21-agent-initialization-skill.md
# `agent-creation` renamed to `agent-initialization` - **Date:** 2026-08-21 - **Type:** refactor - **Scope:** `skills`, `web`, `docs` - **PR:** [#400](https://github.com/Prism-Shadow/penguin-harness/pull/400) - **Breaking:** yes — the library Skill `agent-creation` no longer exists under that name [中文版](2026-08-21-agent-initialization-skill.zh.md) The Agent Tuning group's `agent-creation` Skill is now **`agent-initialization`**. The name says what it does: it initializes an Agent's settings — AGENTS.md, identity metadata, and the Skills that Agent needs — rather than creating an Agent, which the product does elsewhere. The Skill's own copy follows the name (`# Agent Initialization`, a rewritten description and both short descriptions), and its `version` moves to 8. Behaviour is unchanged: same steps, same output, same files written. Every live reference moves with it — the group registry, the Skills library table in the docs, the draft screen's example task that pins it, the SDK Skill's cross-reference, the READMEs and the SDK example. Published blog posts and the frozen changelog entries of released versions keep the old name: they record what was true when they were written. ## Compatibility An Agent that already installed `agent-creation` keeps its copy at `agent_state/skills/agent-creation/`. That copy still works — it is a self-contained directory — but it no longer corresponds to any library Skill, so it will never be updated again, and the library will offer `agent-initialization` alongside it as uninstalled. **Installing the new one leaves both on disk, doing the same job under two names.** There is no automatic migration, by decision. To move over: install `agent-initialization` from the Skill library, then delete `agent-creation` from the Agent's Skills tab. Anything that pins the Skill by name — a prompt saying "use the agent-creation skill", a saved shortcut, an `AGENTS.md` that names it — needs the new name; a stale pin resolves to nothing rather than failing loudly.
Use when developing PenguinHarness itself — changing packages/{core,server,web,cli,desktop,landing,docs,skills}, the built-in model catalog, the installers or the release workflow; writing or auditing changelog entries; writing a blog post or capturing release screenshots; deciding what to do about data already on disk; or auditing prose that reads like a leaked authoring session. Covers the two-repo symlink layout, the CI-parity verification chain, the record-and-ship contract, where blog media is hosted, and the seams that are intentional.
Use when changing the PenguinHarness Web App (`packages/web`) — adding or restyling any UI, picking a status colour, adding an icon, laying out a row or a form field, writing user-facing copy, or building a popup. Covers the semantic tone tokens, the icon size/stroke/gap scale, the semantic-versus-formatting rule for explanatory text, the two-dictionary i18n contract, and the portal-panel pattern with its Esc and scroll caveats.
Run one specified Test Agent on one specified Benchmark Case exactly once, privately score that execution, and return one protocol result.
Initialize an Agent's settings from a user requirement by writing AGENTS.md, setting identity metadata, and installing only needed Skills.
Improve an Agent State through versioned scores and score-linked Traces from a frozen Benchmark.