Wrap-up — close it out¶
The explicit close-out. Confirm what's live, regenerate the shareable overview,
and name the next 1–2 upgrades so the founder leaves with a clear roadmap. The
./my-agent/ folder keeps working after the session ends.
Workflow¶
- Inventory what they own. Tables agent / environment / session / memory / outcome / deployment and the phases completed.
- Regenerate the overview page.
Self-contained
python3 scripts/overview_page.py \ --sheet ./my-agent/build-sheet.json --out-dir ./my-agent \ --status live --loop-shape cron-loop --last-verdict satisfiedagent-overview.html(inline CSS, theme-aware, no external assets) — shareable as-is. - Suggest the next moves. Ranks recorded deferrals (v1 before v2, real-integration first) plus standing hardening (tighten networking, pin the agent version, nest an outcome).
- Finalize. Ensure
NEXT-DIRECTIONS.mdis current (Phase-4 tool), thengoal_state.py advance→phase=done.
Hard rules¶
- Recap what's actually live — read it from the sheet + goal state, never assert primitives that weren't created.
- The overview is single-file — no external assets, so it shares cleanly.
- Every next move names the exact mechanism.
Forcing-question library (recommend + cite)¶
- "Confirm what's live vs still a plan?" Recommend: inventory from the sheet. Cite: this SKILL.
- "Which single upgrade has the highest payoff?" Recommend: the top-ranked v1 deferral. Cite: upgrade_suggester ranking.
- "Is the overview page current?" Recommend: regenerate after any change. Cite: this SKILL.
Tools¶
scripts/primitives_inventory.py— recap every owned primitive.scripts/overview_page.py— regenerate single-file agent-overview.html.scripts/upgrade_suggester.py— next 1–2 upgrades with mechanisms.