Tier-2 Review: one-three-one-rule (methodology cluster)

What we attempted:
We pulled the one-three-one-rule skill from the NousResearch Hermes Agent collection and ran it through the standard harness (D-048). The goal was to verify the skill's executable surface: install it, invoke it with a realistic prompt (e.g., "give me a 1-3-1 on using Postgres vs. DynamoDB"), and observe whether it produced the promised structured output.

What failed:
The harness could not execute the skill at all. Two checks were skipped outright:

• Install — SKILL.md contains no installation command, no package manifest, no dependency list, and no setup instructions.
• Smoke invocation — the skill describes a conceptual framework but provides no CLI, no API, no function signature, and no example invocation syntax. There is nothing to call.

Zero tests passed, zero failed, two skipped. The blocker is structural: this skill is a framework description, not an executable artifact.

What we observed:
Despite the execution failure, the content quality is high. The SKILL.md clearly defines:

• Trigger clarity (4.5) — explicit trigger phrases ("give me options", "1-3-1") and a precise condition (user faces a choice between multiple approaches).
• Output specificity (4.5) — the 1-3-1 format is unambiguous: one problem statement, three options with pros/cons, one recommendation with definition of done and implementation plan.
• Scope precision (4.5) — scoped to technical proposals, architecture decisions, tool selection, refactoring, migration. No scope creep.
• Self-containment (4.5) — the framework is fully described in the markdown; no external references required for understanding.
• Reusability (4.0) — the format is broadly applicable across technical domains, but the lack of an interface limits repeatable programmatic use.

The failure is not a content defect — it's a packaging defect. The skill reads like a prompt template or a coaching guide, not a callable skill. In a human workflow, a user could read this and manually apply the 1-3-1 structure. In an automated harness, there is no hook.

Rating caveat:
The composite score of 4.4 / 5.0 is theoretical. It reflects the quality of the written framework, not its demonstrated behavior. Until the skill is re-packaged with at least one of the following — a documented CLI, a Python entry point, a JSON schema for input/output, or a template file that can be rendered — the score cannot be validated through execution. A physical re-run after adding an executable surface is required to confirm the ratings.

Verdict on value:
Yes, the skill remains valuable in principle. The 1-3-1 format is a proven decision-making structure, and the trigger definitions are unusually precise. The core intellectual content is sound and would likely score well once wrapped in a minimal interface (e.g., a run.py that takes a problem statement and returns a markdown-formatted 1-3-1). The failure is fixable with modest effort — roughly 15–30 minutes of authoring an invocation script and a one-line install instruction. Until then, treat the 4.4 as an upper bound for content quality, not a verified performance metric.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-08-07.

Overall: broken. 0 tests passed, 0 partial, 0 failed, 2 skipped; key blocker: SKILL.md provides no executable interface or installation instructions.

Test	Status	Notes
install	skipped	SKILL.md does not document an installation command; no install instructions provided.
smoke-invocation	partial	SKILL.md describes a conceptual framework but lacks a concrete CLI or API; minimal invocation cannot be executed as a command.