audit: worktree-audit.sh rg skips sessions.db as binary #24

port: skills/poteto-mode/scripts/worktree-audit.sh L64–66 and L106–108 at 9309f40 (after #23):

# CLI session store: ~/.local/share/devin/cli/sessions.db
# ($XDG_DATA_HOME or DEVIN_SESSIONS_DIR when set). No per-OS branch.
transcripts="${DEVIN_SESSIONS_DIR:-${XDG_DATA_HOME:-$HOME/.local/share}/devin/cli}"
...
	if [ -d "$transcripts" ]; then
		f=$(rg -l -e "${wt}/" -e "${wt}\"" "$transcripts" 2>/dev/null \

upstream pstack/skills/poteto-mode/scripts/worktree-audit.sh L64–66 and L106–108 at 88d4751 (pstack 0.15.29):

# Transcripts dir: ~/.cursor/projects/<slugified-repo-path>/agent-transcripts.
slug=$(printf '%s' "$main_wt" | sed 's#^/##; s#/#-#g')
transcripts="$HOME/.cursor/projects/$slug/agent-transcripts"
...
	if [ -d "$transcripts" ]; then
		f=$(rg -l -e "${wt}/" -e "${wt}\"" "$transcripts" 2>/dev/null \

why it's a port divergence: upstream greps a directory of .jsonl text transcripts. The port now points transcripts at the Devin CLI dir, where session history is the SQLite file sessions.db. When rg recurses into a directory, it stops reading a file at the first NUL byte, and the SQLite header has NUL bytes, so sessions.db never matches. Any hit comes from logs/*.log instead, so LAST_CHAT reads - or a log date. Reproduced with ripgrep 14.1.1 on a one-row sqlite file: rg -l -e '<path>/' . exits 1, and rg -l -a -e '<path>/' . lists ./sessions.db. The #23 harness proof saw the same on a real ~/.local/share/devin/cli.

(The stat -f '%m %N' / date -r on L108–110 is macOS-only, but upstream has the same lines, so it's an upstream bug and not part of this issue.)

smallest fix: L107: add -a so rg searches the binary store:

		f=$(rg -l -a -e "${wt}/" -e "${wt}\"" "$transcripts" 2>/dev/null \

found by the whole-port audit against upstream 88d4751 (pstack 0.15.29). fix with Closes #24.