Tier-2 Review: inbox-setup

What We Attempted

We attempted to validate the inbox-setup skill from the SkillsMP marketplace, which promises to build a personalized inbox triage knowledge base through an interactive interview process. The skill is designed as a one-time setup that interviews users about their email patterns, business context, reply style, and priorities, then generates knowledge base files for a companion inbox-triage skill.

What Failed

The skill failed to run cleanly in our test harness. Of three test categories, zero passed, one was partial, and one failed outright.

Install (fail): The skill has no install command documented in its SKILL.md. It is not a pip package and does not follow standard installation patterns. The skill is described as a "one-time interactive setup" that runs within a Claude CLI environment, but no mechanism for automated installation was provided.

Smoke-invocation (partial): When invoked, the skill triggers an interactive interview that requires real-time user input and a Claude API key. There is no automated test path—the skill demands human interaction at every step, with no mock mode or test harness support.

What We Observed

The core blocker is architectural: this skill is fundamentally interactive and not designed for automated testing. It operates as a conversational setup wizard, not as a programmatic tool. The skill's design philosophy—one question at a time, forcing format where possible, dependency-ordered, with explanations for each question—makes it excellent for human use but impossible for automated validation.

The skill's dependencies are claude-cli and python>=3.8, but no concrete install steps or package requirements are listed. The skill is meant to be run once and then left alone, which means it has no reusability in the traditional sense (a factor reflected in its D5 reusability score of 3.5).

Rating Assessment

The composite score of 4.5/5.0 is theoretical until a physical re-run in a proper Claude CLI environment can confirm the skill's behavior. The dimension scores—particularly trigger clarity (5.0) and output specificity (4.5)—are based on the skill's documentation, which is well-structured and clear. However, without a working installation and testable invocation, these scores cannot be empirically confirmed.

Value in Principle

Despite the test failures, this skill still appears valuable in principle for its intended use case. The concept of an interactive setup wizard that builds a personalized email triage system is sound and addresses a real pain point for knowledge workers. The skill's thorough approach to interviewing the user about business context, pricing, priorities, and reply style suggests it would produce genuinely useful output.

The skill is not broken—it is simply not testable in an automated harness. For a human user with a Claude CLI environment and a willingness to complete a structured interview, this skill would likely deliver on its promise. The test failures are a limitation of our test methodology, not necessarily a reflection of the skill's quality.

Recommendation: This skill should be re-tested in a live Claude CLI environment with human interaction. Until then, treat the 4.5 rating as a best-case estimate based on documentation quality, not empirical validation.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-06-13.

Overall: broken. 0 tests passed, 1 partial, 1 failed; key blocker: skill is interactive and not installable via pip.

Inferred dependencies: claude-cli, python>=3.8.

Test	Status	Notes
install	fail	No install command documented in SKILL.md; skill is a one-time interactive setup, not a pip package.
smoke-invocation	partial	Invocation triggers interactive interview; requires user input and Claude API key; no automated test possible without mocking.