Tugboat v0.1.1

Tugboat now treats concise communication as a core behavior, not a demo-only style choice.

• Deep work no longer implies long replies.
• Progress updates default to change, meaning, and next action.
• Details expand when risk, uncertainty, permission, or a user decision requires them.
• E16 adds regression coverage for communication overload.
• English and Traditional Chinese demos now reflect the concise default.

Safety invariant: brevity never hides setbacks, material evidence, uncertainty, or permission boundaries.

Tugboat v0.1.0

Tugboat v0.1.0 is the first public release of an anxiety-aware, evidence-driven Agent Skill for stalled or high-stakes work.

What is included

• A collaboration workflow that turns perspective-taking into investigation, execution, and validation
• Protection for the user's stated goal while methods and intermediate paths remain flexible
• Outcome, causal, and directional definitions of credible progress
• Confidence reporting tied to visible evidence rather than reassurance
• Bounded persistence, path-switching rules, and normal safety and permission limits
• De-identified evaluation cases and bilingual documentation
• Installation guidance for Codex, Claude Code, GitHub Copilot CLI, Pi, Gemini CLI, and Grok Build

Install

gh skill install grgy078033/tugboat

Review the skill before installing:

gh skill preview grgy078033/tugboat tugboat/SKILL.md

Tugboat does not diagnose, provide therapy, or guarantee an outcome that the evidence cannot support.

Full changelog: https://github.com/grgy078033/tugboat/commits/v0.1.0