- Package:
skills/language-learning/ - Invocation type: user-invoked only
- Profile:
review-loopagent-skill - Admission status:
PASSunder the low-risk prompt-only fast track - Stable-release boundary: v0.1.1 contains five packages and does not contain
language-learning
| Area | Result | Evidence boundary |
|---|---|---|
| Source | PASS | Original first-party design; no copied third-party code, scripts, or assets. |
| Structure | PASS | 33 contract assertions; 931 collection-discovery assertions across all seven packages; valid frontmatter and resolved links. |
| Invocation | PASS | Claude disable-model-invocation: true and Codex allow_implicit_invocation: false; user-invoked only; non-trigger returned NOT_INVOKED. |
| Fresh-copy install | PASS | Isolated copy contained only language-learning, no source checkout, identical file set, zero SHA-256 mismatches, installed contract tests 33 assertions PASS; host install byte-identical and discovered in the host skills root. |
| Behavior | PASS | Fresh agents produced a routed flashcards set and a beginner-defaulted daily lesson without re-asking language/level/mode. |
| Documentation synchronization | PASS | Collection discovery, catalog, bilingual guides, maintenance baseline, and changelog agree on seven packages on main versus five in stable v0.1.1. |
| Independent review | PASS | A fresh final fast-track Evaluator confirmed eligibility, reproduced the evidence, verified all nine acceptance criteria, and returned PASS. The Evaluator raised one Low-severity evidence-accuracy observation, resolved in the Producer record. |
Full records are under review-loop/.
Local-source and host-install evidence is admission evidence, not proof of a
released install command. A pinned language-learning command must wait for the
next published tag and fresh released-repository verification.
Original first-party design. No upstream Skill code or prompt text was copied.