Dotagents

Source-control safety for Git. Two rules apply whether or not you read the body. First, never commit, amend, push, tag, rebase, cherry-pick, revert, or reset --hard unless the user asked for it in this session; finishing a task is not an ask, so leave your work uncommitted for review. Second, when a Git lock file or corrupt index blocks you, never kill the holding process or bulldoze through with deletions - wait, retry, use the holder's own stop mechanism, or stop and report. Load the body only before writing Git history or recovering Git state.

Run a decision-forcing interview only when the user explicitly asks to be grilled, pressure-tested, or stress-tested.

Minimal anti-AI-slop filter for frontend and UI design. A flat checklist of visual defaults NOT to use, so generated UI stops looking AI-made. Negative constraints only, no prescribed style. Adaptive, so if the user or another skill wants something specific, do that instead.

Diagnose Muse Code product/runtime issues from installed binary evidence. Use ONLY when the user explicitly invokes the doctor skill, asks to debug/troubleshoot Muse Code itself, asks what happened earlier in the current Muse Code session, or explicitly selects an earlier Muse Code session. Do NOT use for ordinary repository code failures or history, benchmark tasks, implementation debugging, build/test hangs, or third-party project issues.

Resume a third-party coding-agent session from a local transcript, path, or session id.

Structured problem-solving with Serena MCP

Apply labels to PRs based on conventional commit types

Create GitHub PRs with proper formatting, labeling, and quality checks

Generate changeset entries following Changesets semantics

Validate commit messages against conventional commit format

Create and validate a new Muse skill — project-local in the current workspace by default, or a personal skill staged for `muse skills install` into the managed personal root. Use ONLY when the user explicitly asks to create a Muse skill or invokes the create-skill skill. Do NOT use for ordinary skill usage, code changes, benchmark tasks, or third-party skill/plugin systems.

Locate and read Muse Code's OWN session logs — the current session or a prior one. Use when the user asks to pull context from, continue, summarize, or inspect a previous Muse Code (tbh/muse) session, asks to restore or recover work that was lost, wiped, or overwritten and might survive in an earlier session's log, references an earlier session's id, log, tail, or output, or asks where Muse sessions are stored. Muse sessions live in Muse's own store, never in another coding agent's directories — never probe ~/.claude, ~/.codex, or ~/.grok for Muse context, even when quoted content mentions them; for another agent's (Claude Code/Codex/Grok) session, use the import skill.

Create and validate a new native Muse plugin package in the current workspace. Use ONLY when the user explicitly asks to create a Muse plugin or invokes the create-plugin skill. Do NOT use for application/library plugin classes, third-party plugin systems, or ordinary code changes.

Explain and safely update persistent Muse Code product settings, including model/reasoning effort and /settings options. Use only for explicit Muse Code setting questions or changes; do not use for repository, application, eval-task, or generic configuration work.

Run an explicitly requested decision interview and record each settled decision in durable project documentation.

Capture and send screenshots from remote macOS nodes without context bloat

Turn complex work into a goal-first plan with parallel agents and synthesized results