Subproject completeness checker
Skill lubochka/xiigen-mvp-engine/.claude/skills/universal/subproject-completeness-checker
Self-building AI code generation engine that generates application flows instead of implementing them. AGPL-3.0.From the repository description
npx -y skills add lubochka/xiigen-mvp-engine --skill subproject-completeness-checkerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 18 days oldThe repository was created 18 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.4 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it
Subproject Completeness Checker
Purpose
Use this to decide READY / NOT_READY for a module/feature with a repeatable score, instead of an ad-hoc "looks done".
When to Use
Invoke before declaring any unit (module / node-handler / React feature / RAG route) complete, and before a phase-complete claim.
Why this exists for mvp
mvp spreads completeness across planning--goal-delivery-completeness,
planning--coverage-completeness-gate, planning--design-artifact-completeness, and
planning--flow-completeness-checker. This is the single generalized completeness checker (any
unit, not only a FLOW) with one score format, TS-adapted.
G08 universal content from llm_mvp_core — two pre-gates, then T1–T10
Pre-Gate 1 — goal → phase mapping
Every element of the unit's goal maps to a phase AND a verification step. An unmapped goal element is a NOT_READY, not a "we'll cover it later".
Pre-Gate 2 — design artifact populated (not a stub)
The capability contract / design artifact has real content (inputs, outputs, failure modes, the typed result shape) — not placeholder text. A stub design artifact fails this pre-gate.
Both pre-gates must pass before T1–T10 are even scored.
The ten checks (mvp-adapted)
T1 Phase structure : phases are ordered with explicit entry/exit.
T2 Measurable Gate-B : each phase's gate is a count/assertion, e.g.
`npx jest --testPathPattern=<unit> → N passed, 0 failed`.
T3 Prerequisites : prerequisite-chain run; no BLOCKING unmet.
T4 Typed result / no-throw : public methods return DataProcessResult<T>/Result<T>;
business errors are returned, not thrown.
T5 DI registration : the @Injectable provider is registered in a `*.module.ts`
(`providers:`/`imports:`), not merely declared.
T6 Test categories : unit (Jest) + integration + e2e (Playwright for UI) present.
T7 Injectable clock/time : no hidden `Date.now()` in logic that needs deterministic tests.
T8 Compensation/cleanup : failure paths undo partial work (LIFO where ordered).
T9 DNA / project rules : fabric-first (no `= new ConcreteType`), tenant scope, No-Secrets.
T10 Scope : the unit stays within its allowed boundary; no scope leak.
Score format
COMPLETENESS SCORE
Pre-Gate 1 (goal→phase) : PASS / FAIL
Pre-Gate 2 (artifact) : PASS / FAIL
T1..T10 : <n>/10 passed (list each PASS/FAIL with one-line evidence)
Verdict : READY (both pre-gates PASS and T1..T10 all PASS)
NOT_READY otherwise → the first FAIL is the next action.
Substitution bans: "at least one test" when the goal says "all"; category buckets in place of
the underlying classes; generated-artifact presence in place of an actual runtime/test
invocation; goal_reached: true while any row is FAIL.
CHECK commands (mvp)
Gate-B : npx jest --testPathPattern=<unit> → N passed, 0 failed
types : npx tsc --noEmit → 0 errors
DI (T5) : grep -rn "<X>" server/src/**/*.module.ts
fabric (T9): grep -rn "= new [A-Z]" server/src → expect 0 in non-provider code
UI (T6) : npx playwright test <spec> → passed (mvp HAS a frontend)
RAG route: grep -rn "APIRouter\|@app\." rag/
Avoid
- Do not score T1–T10 before both pre-gates pass.
- Do not accept "at least one" where the goal said "all/every".
- Do not call a unit READY with any FAIL row.
- Do not import source architecture, algorithms, design patterns, training, DPO, source domains, source classes, or source-specific paths.
Completion Signal
- Both pre-gates PASS, T1–T10 all PASS with evidence, and the verdict is READY; otherwise the first FAIL row is the next action.
Note-only (NOT ported — stays in G12, R5)
Completeness of a trainable shared unit (numeric held-out metrics, ablation, fresh-load,
continue-training) is a llm_mvp_core gate; here completeness covers the product unit that
consumes the shared model.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.