Keeping the portfolio honestReview

A portfolio lies by omission: a project stops moving and nothing tells you. Today is about building the cheap tripwire that catches staleness before it costs you two weeks — the same failure that just hit your Campaign Comms PRD.

A dashboard shows you what's there; it does not show you what's *missing*. PDB/projects.md faithfully lists every active project — but a project that hasn't moved in two weeks looks identical to one you touched this morning. The rot is invisible because the absence of activity produces no event, and you can only react to events.

The fix is to convert absence into a signal. Instead of waiting to *notice* that something went quiet, you periodically scan every project's last-touched time and emit a loud artifact for anything past a threshold. This is the same principle as your automation-hygiene rule — 'verify the side-effect, not the trigger' — applied to your own attention: don't trust that you'll remember a project, make the system surface the ones you've forgotten.

In your workspace the raw material is already there: every project has a file under PDB/data/projects/ and a workspace under JARBUS/projects/<slug>/, and the filesystem stamps each with a modification time. A staleness sweep is just: list them, ask the OS when each was last written, subtract from today, flag the laggards. No database, no daemon — one command answers 'what have I been ignoring?'

Worked example

Find every active-project file not touched in 10+ days, newest-first so the freshest float to the bottom:

cd ~/Claude/PDB/data/projects
find . -name '*.md' -mtime +10 -print0 | xargs -0 ls -lt

# Same idea over the JARBUS work itself (where the real edits land):
cd ~/Claude/JARBUS/projects
find . -maxdepth 1 -type d -mtime +10 -exec ls -ld {} \;

• find . -name '*.md' walks every markdown file under the projects ledger.
• -mtime +10 means 'last modified more than 10 days ago' — the staleness threshold. Drop to +7 to be stricter.
• -print0 | xargs -0 pairs filenames safely even if they contain spaces (your slugs do).
• ls -lt sorts the survivors by time so the most-stale sit at the top of the list — that's your worklist.
• The second block checks the JARBUS workspace folders, because a project's PDB metadata file can be fresh while the actual asset/ and drafts/ have gone cold — that gap is exactly the lie.

Gotchas

• mtime lies if a script rewrites the file without real progress — e.g., a metadata bump. Cross-check against the git log of the actual workspace, not just the ledger file's timestamp.
• A 10-day threshold will flag legitimately-paused projects too. Staleness detection finds *candidates* for attention, not verdicts — you still make the call.
• Don't automate the shelving. A tripwire that auto-archives is how you lose a project you were about to return to. Surface loudly, decide manually — same as your 🚨 Things tasks.