Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →jpvarbed/goal-to-done
Steward one substantial goal from a rough request to verified completion - clarify consequential decisions, compile a verifiable goal brief, walk the closed seven-route cascade to the smallest fit (four outcome routes total-tdd, wayfinder, direct, to-tickets; three interstitial outcomes grilling pass, human pause, no-work receipt), drive the dependency-ready ticket frontier through existing harness and tracker mechanisms, re-plan when evidence invalidates the graph, and finish with integration-l
Passed 13 of 19 test cases with the skill, and 4 of 19 without it.
and 18 more in the full report
12 fixed5 passed either way7 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
Wayfinder finished investigation tickets for goal brief v1, proving that external API rate limits require splitting execution into 2 distinct units. Does this first exit from wayfinder require incrementing `[graph-revision]` from 1 to 2 and running section 6 reconciliation before dispatching work?
| Checked | Without skill | With skill |
|---|---|---|
| Re-routes cleared multi-unit work without incrementing graph-revision. | Missed | Passed |
| Keeps grilling predicate closed during post-wayfinder re-routing. | Missed | Passed |
The instructions your model reads when the skill loads.
Steward one goal from rough request to verified DONE. SPEC.md in this
directory is the normative contract; when in doubt, it wins. You compose
existing mechanisms — tracker skill, harness gate/build/verify/judge, TDD,
review council, delegation — and never build a controller, scheduler, daemon,
state store, tracker adapter, or second ticket runner. The tests/ and
evals/ here validate this skill's delivery; never run them to drive real
work.
Before stewarding, confirm the tools this skill composes are actually available — a missing mechanism is a stop, not something to reimplement:
linear auth whoami >/dev/null || echo "tracker CLI unavailable - authenticate or stop"
harness --help >/dev/null || echo "harness unavailable - per-unit gating impossible"
council --preflight-only --workspace . # reviewer seats ready? fail-closed
Check the parent tracker issue for a current non-stale steward claim. If one exists, stop — you are not the steward. Otherwise claim the parent through the host's task controls and the tracker comment:
linear issue view GOAL-123 # look for a current [claim] comment
linear issue comment add GOAL-123 --body "[claim] steward=<agent> session=<id> started=<iso8601>"
Only the steward mutates the graph or claims child tickets; workers execute steward-issued units and never self-claim.
Compile the overall goal with goal-spec into one binary destination. On the
parent issue record: goal-brief version 1, the SHA-256 of the exact UTF-8
goal-brief block, and graph revision 1 ([graph-revision] 1 +
test cases passed, vs 4 / 19 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.