get2knowio / get2knowio/maverick
Spec Kit path has never been validated end-to-end
- Dominant language
- Python
- Stars
- 4
- Forks
- 0
- Avg merge
- 17h 37m
- Merged PRs (30d)
- 7
Description
Follow-up to #135 subtask 5.
That subtask specified the legacy pipeline — `plan generate` -> `refuel`
-> `fly` — and running it found seven real bugs (fixed across `5f2d082`,
`ceff42a`, `c4fec21`).
But **all seven live in `workflows/refuel_maverick/`**, and none of them
exist on the Spec Kit path:
| bug | legacy `refuel` | `refuel --speckit` |
| --- | --- | --- |
| constraint criteria burn 3 fix rounds | yes | no validator, no fix loop |
| SC ref extraction | yes | no |
| structlog leak | yes | no |
| split plan directories | yes | no |
| gap count without text | yes | no |
| briefing failure fatal | yes | no briefings |
| briefings cached too late | yes | no |
`grep -rn "contrarian\|briefing" src/maverick/workflows/refuel_speckit/`
returns nothing; the only model call on that path is the optional
`--enrich` step.
So the walkthrough validated a path that is no longer the recommended
one, and validated **nothing** of:
* `maverick spec` (the headless Spec Kit chain, 050)
* `refuel --speckit` deterministic ingestion
* the assumption ledger and clarify interception (049)
* `maverick reconcile` (051)
* the conditional land gate (052)
* the review console JSON verbs (053)
## Proposed run
On a box with a provider authenticated, against sample-maverick-project
(now reset and working — see its `CLAUDE.md` for the baseline
procedure):
```
maverick spec greet-cli-v2 --from-prd docs/greet-cli-prd.md
maverick refuel greet-cli-v2 --speckit
maverick fly --epic --max-beads 3
maverick review --list --json
maverick land --status --json
```
Deterministic ingestion should make this far cheaper than the ~50-minute
AI decompositions the legacy path required. File a follow-up bug per
regression rather than fixing in-line.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Contributor guide
Research direction
Start with sample-maverick-project/CLAUDE.md for the baseline procedure, then compare workflows/refuel_maverick/ with src/maverick/workflows/refuel_speckit/ and run the listed maverick spec, refuel, fly, review, and land commands. Confirm the Spec Kit chain, deterministic ingestion, assumption handling, reconciliation, land gate, and review JSON behavior; file a separate follow-up bug for each regression found.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- cli, testing-qa
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100