The ~430-line body avoids explaining concepts Claude already knows and is mostly prescriptive, but carries redundant philosophical framing (e.g., "A spec is a launching pad, not a script to follow" restating the Core Philosophy bullets) and prose that could be tightened. It is efficient but not lean enough that every token earns its place, so it lands at 2 rather than 3 and is well above the padded score-1 anchor.

It provides concrete, copy-paste-ready templates for every section, a concrete naming convention ("specs/YYYYMMDDThhmmss-feature-name.md"), decision-class tables, and verbatim before/after call-site examples. Per the instruction-only scoring note, the absence of executable code is not penalized because the guidance is highly actionable.

The spec-writing process is sequenced via the document structure, gated by the one-sentence-test ("the spec is not ready"), and includes an explicit Build/Prove/Remove wave ordering with a verification checkpoint ("Do not schedule deletion before verification passes") plus checkbox checklists. This matches the score-3 anchor of clear sequence with explicit validation steps and feedback loops.

The References section gives condition-based, one-level-deep loading (decision-hygiene.md, verified to exist with no nested references) and the main file keeps core templates inline while pushing deeper failure-mode detail to the reference. It matches the score-3 anchor of a clear overview with well-signaled one-level-deep references and easy navigation.

It names one concrete action ("Write technical specs that let agents implement autonomously") but does not list multiple specific actions, matching the score-2 anchor that names a domain and some actions without being comprehensive. It is not vague (ruling out 1) and does not enumerate multiple distinct actions (ruling out 3).

It explicitly answers both what ("Write technical specs that let agents implement autonomously") and when (the "Use for ..." trigger phrases), matching the score-3 anchor that clearly answers both with explicit triggers.

The "Use for 'write a spec', 'plan this feature', 'create a planning doc'" clause gives natural phrases a user would actually say, with good coverage of common variations. It is far above the jargon-only or single-keyword anchors for 1 and 2.

The specification-writing niche with distinct spec/planning-doc triggers is clearly distinguishable and unlikely to fire for unrelated skills, matching the score-3 anchor. While 'plan this feature' has slight overlap with general planning, the spec-writing framing keeps it distinct, so it stays above 2.