deva-clean

# deva-clean: mount-safe workspace disk cleanup A deva workspace mixes two kinds of directories that look identical in ls: bind mounts declared in `.deva` (live host repos) and local dirs created during work (git worktrees, node_modules, package stores). rm -rf on the wrong one destroys a live repo on the host. And because the workspace root is itself a host mount, local dirs still consume host disk - usually the disk that is actually full. Two passes. Pass one is deterministic and scripted; pass two executes only what got approved. ## Pass one: triage Run the bundled script - read-only, emits the tiered plan, never deletes: scripts/triage.sh [workspace] [--fetch] It maps mounts from /proc/mounts (ground truth - .deva drifts), shows which disk is under pressure, walks the non-mount dirs, gives every candidate four verdicts (type, dirty, pushed, freshness), and finds stale worktree metadata in the mounted parent repos. Verdict semantics live in the script header; two encode traps worth knowing: - The pushed test is live ls-remote SHA comparison plus ancestry in the default branch - never upstream config. A branch created off origin/develop shows "ahead 1" while never pushed itself. - MERGED is monotonic, provable even against stale refs. Absence is not: a branch absent on remote with stale refs is UNVERIFIED, not safe. Rerun with --fetch for a definitive verdict. Then layer the judgment the script refuses to automate: - Pressure: container overlay genuinely full (rare - on shared-VM runtimes overlay df reports the whole VM disk) means container caches (~/.cache, package stores) are the target, not the workspace. - Stray files: the script walks dirs only; shell-typo artifacts like "&1" are yours to spot. - Promotions: KEEP entries flagged MODIFIED TRACKED FILES move to tier 2 only when you can name every dirty file as a generated artifact (lockfiles, test output) and say so in the plan. - Demotions: tier 1/2 entries that are functionally live - symlink targets, open-PR checkouts - move to keep. - Keepers: offer cache-prune (node_modules, dist, coverage) as the low-risk alternative; for LOCAL-ONLY work, say exactly what would be lost and suggest pushing it. Done when every candidate sits in a tier you can defend. Present the plan and stop - pass two starts only from explicit approval, tier by tier. ## Pass two: execute approved tiers - Linked worktrees: `git -C <parent> worktree remove <path>` is the only removal path. Its refusal on a dirty tree is the safety net, so no --force - and no rm -rf, which bypasses the net and leaves stale metadata behind. - Tier 2 trees: first delete the named generated files (restore tracked ones with `git checkout --`), so the tree is genuinely clean and the safety net stays armed. - Stale metadata: `git -C <parent> worktree prune`. - Standalone repos (pushed, clean) and plain dirs: rm -rf. Safe by construction: `git worktree remove` deletes the checkout only; branch refs live in the parent repo's .git and survive. Committed work is never lost - only uncommitted files can be. The metadata edit inside the parent's .git is normal bookkeeping, not touching the mount. Done when df shows the reclaim and a triage rerun reports empty tiers.