docs(agent-guide): document when to use muse vs muse-dev (#185)
Gabriel noticed I kept using plain `muse` for ordinary VCS operations even while Phases 1-3 stood up the muse/muse-dev split, and asked for it to be documented so future agents don't have to infer the rule from behavior.
Adds a dedicated section: `muse` (stable, installer venv) is the default for everything, including work inside ~/ecosystem/muse itself — ordinary content work (docs, tests, features, commit/merge/push) is not "developing muse's CLI," it's just using a VCS tool. `muse-dev` (editable, live source) is only for exercising a CLI/engine code change you just made to muse itself. Also documents the Phase 3 guard rail's existence and rationale so agents that hit it understand why, rather than reflexively setting MUSE_DEV_ALLOW_CANONICAL=1 to make an error go away.
Also added a condensed version of the same rule to the sub-agent boilerplate block, since sub-agents only ever see that snippet, not the full doc.
Semantic Changes
3 symbols
Files Changed
~1
1164 in snapshot
0 comments
muse hub commit comment sha256:055a23adbc9f95a2b8db00703e661d8dcce5be185c5eff1282f84e47e9adb59d --body "your comment"
No comments yet. Be the first to start the discussion.