follow-through
Brings selected alerts, household follow-ups, research choices and colleague promises back to their sources. It gives each loose end an evidenced state and next action, keeps stale observations visible, and verifies the authorised outcome before calling it complete.
/plugin install follow-through@fledgeling-pluginsNeeds the marketplace added first — how to do that.
Reach for it when
Reconcile a bounded set of operational alerts, household follow-ups, research decisions or colleague commitments into evidenced next actions. Use when loose ends span these areas; route a single area to its dedicated skill. Does not start a monitor or contact anyone merely because an item is open.
What ships with it
- No scripts — prose only
- 3 reference files
- Measured evals
Open items tend to spread across mail, tracking pages, research tabs and conversations. This plugin brings the selected set back to its sources, works out what still needs doing, and keeps the next action attached to the evidence. Old screen activity can point to the source; it cannot prove today's balance, delivery or completed promise.
Install from this repository's marketplace:
/plugin install follow-through@fledgeling-plugins
| Invocation | Use it for |
|---|---|
/follow-through:follow-through Review these loose ends | A bounded mixed review |
/follow-through:ops-triage Review selected project alerts and credits | Operational errors, cost and credit deadlines |
/follow-through:household-follow-up Check these two shipments | Deliveries, returns and household administration |
/follow-through:decision-brief Decide a storage approach from this research | Product/tool research with a decision to make |
/follow-through:commitments Reconcile promises in these conversations | Colleague commitments and handoffs |
For repeatable UI checks use test-campaign:ui-regression; for already authorised
agent continuation use flagship:fleet-console. Referrals are checked against the
receiving runtime. Missing connectors or skills remain visible as gaps.
Creating these skills grants no authority to start monitors, send messages, buy, book, redeem credits, change accounts or release software. A review produces evidence and next actions. When the user authorises execution, the skill carries that bounded work through and verifies the requested outcome.
An optional read-only helper validates a private JSON ledger and prints a summary:
python3 plugins/follow-through/scripts/ledger.py check /path/to/private-ledger.json
python3 plugins/follow-through/scripts/ledger.py summary /path/to/private-ledger.json
python3 -m unittest discover -s plugins/follow-through/tests -v
Its format is bundled in the router skill. No daemon, connector credentials, settings changes or personal data ship with this plugin. See EVALS.md for what was tested and what remains unproved.