Dobeu-tech-eco

dt-fullauto

"dpp-full with autopilot forced on, looping until a check-in reports nothing outstanding. Use when the user says fullauto, run everything, or finish the end state unattended."

Dobeu-tech-eco 0 Updated 1w ago
GitHub

Install

npx skillscat add dobeu-tech-eco/dts2-marketplace/dt-fullauto

Install via the SkillsCat registry.

SKILL.md

Cursor procedure

Find the plugin root: the nearest ancestor of this file that contains .cursor-plugin/plugin.json. The examples below say $PLUGIN for that directory.

Run python3 "$PLUGIN/scripts/dt_gate.py" for every gate. Exit 2 from scaffold, preflight, or approve stops the command. Exit 2 from bridge that prints BLOCKED unsupported-runtime-script is a precise blocker for that one step: record it and do not imitate the script. The interview, approve, and scaffold still run.

--yes, --ship, and --scaffold-only are rejected. A JSON key with the same name as a kind is not authorization. approve records the sha256 of the file the user accepted, and later gates require that digest.

Phase files for this command are in $PLUGIN/skills/dt-fullauto/.

Merges, fleet broadcast, hook install, the headless SDK, and paid provider calls are CURSOR_UNSUPPORTED even after a permission phrase is recorded.

Cursor edition: run scripts/dt_gate.py from this plugin for every gate. Do not read tokens or credential files. Do not merge, broadcast, install hooks, or spend a provider API. Carry out file work with Read, Edit, Grep, Glob, and git.

fullauto — dpp-full under autopilot, to the end

Run skills/dpp/SKILL.md as dpp-full with autopilot on for every lane; the head re-runs /dt-check-in after each son reports done or incomplete and before the next dispatch, so new items join the candidates. Each pass first runs lanes.py loop <stamp>; when it returns stop: true (max_fullauto_loops reached) the run parks as incomplete with the ledger written and the final report lists what is left.
Plan with --full --fullauto --autopilot (dpp section 5.5): under fullauto_freeze_scope: true the item set approved at the first plan is frozen and new check-in items wait in the pending list. Stop when lanes.py stop-check <stamp> --from-checkin <check-in.json> --base <branch> (the end state read from the fetched origin/<branch>; items tied to a ledgered blocked PR do not count as outstanding) says a check-in run found nothing outstanding, with each lane done, or blocked and listed in the task ledger (autopilot §5a).
Every pass also follows dpp §5 and §6: python3 "$PLUGIN/scripts/dt_gate.py" bridge lanes.py reconcile <stamp> before a PR is created, merged or a lane resumed, and python3 "$PLUGIN/scripts/dt_gate.py" bridge lanes.py transition for every phase move. When the loop stops, for any reason, end with python3 "$PLUGIN/scripts/dt_gate.py" bridge lanes.py shutdown <stamp> and print python3 "$PLUGIN/scripts/dt_gate.py" bridge lanes.py report <stamp>.
Stop early only on an autopilot §4 hard stop or a lane the council could not clear (autopilot §1b); the final report then lists the remaining items. Unattended outer loop (autonomy: unattended in .harness/config.yaml lifts the max_fullauto_loops cap; max_fullauto_loops: 3 stays the default): set /goal to "all ready items are done or parked, or loop_guard.py returned a stall verdict or STOP" and run /loop over .claude/loop.md, copied from templates/loop.md. Each pass runs python3 "$PLUGIN/scripts/dt_gate.py" bridge loop_guard.py check (exit 2 prints STOP), then loop_guard.py sweep --release, picks the next ready item, does it as one round, loop_guard.py parks an item that needs a person, and ends with loop_guard.py record --progress "<open items> | <merged PRs>" --signature "<failure>". A stall verdict (no progress for loop.no_progress_limit passes, or one failure signature three times in a row) or a stop runs loop_guard.py notify, which opens or updates the one "dts2: needs you" issue and writes a local note, then the run ends with the shutdown and report steps above; merge and policy rules are unchanged.