v1.1.0 - Governed Agent Skill

Product logic

• Chooses direct, compact-panel, or full-panel according to coordination value and risk, not task count.
• Gives one steward final accountability and each specialist one owned artifact with explicit handoffs and checks.
• Adds capability, privacy, budget, approval, failure, artifact ownership, and rollback boundaries.
• Operationalizes a design only when the user requests it and the host can actually support it.

Installable package

The canonical Skill entrypoint is SKILL.md at the repository root. This release includes a small standalone ZIP and its SHA-256 checksum.

Upgrade from v1.0.0

Use the repository root. The old /skill path now contains only a migration notice because that nested package name could not pass the parent-directory naming rule.

Validation

Passed local package validation, official agentskills validation, Yao Meta Skill validation, clean archive installation, and 25/25 repository trigger fixtures. Fixture results do not prove real-world adoption or measured model-quality improvement.

v1.0.0

Initial open-source release of Agent Expert Panel Skill.

Highlights:

• Portable skill/SKILL.md for designing, auditing, and optimizing multi-agent expert-panel workflows.
• Steward, expert-member, capability-preflight, handoff, failure-handling, and decision-trace contracts.
• Bilingual README documentation and methodology notes based on Ye Xiaochai's public Woshipm article.
• Public-safe generated visuals, bilingual HyperFrames demo pages, and rendered MP4 demos with MiniMax CLI background music.
• Security audit, contribution guide, issue template, and test prompts.

Boundaries:

• Does not include WorkBuddy private prompts, private product assets, or non-public implementation details.
• Does not execute subagents by itself; the host agent decides how to spawn or simulate members.

Checks:

• Skill validator passed.
• HyperFrames lint, validate, and inspect passed with 0 layout issues.
• Privacy scan passed for local paths and actual secret values.