Conciseness

The body contains several structurally redundant sections: the 'Information Architecture' bullet list duplicates the file's own headers as a table of contents, 'Quick Commands' repeats the exact commands already embedded in Workflow steps 4-5, and 'Quick Start' overlaps the 'Configuration Examples' section with a near-identical provider JSON block. This is 'mostly efficient but could be tightened' rather than score 4, where over-explanation would be only minor instances — here three whole sections could be removed or merged with no loss of information.

Actionability

Concrete commands (jq, rg, opencode run) and full JSON/markdown templates exist, but the primary copy-paste examples contradict the skill's own references/config-schema.md: SKILL.md shows a top-level 'providers' array with 'baseEnv' and 'default: true' model entries and a 'permissions' object with 'filesystem'/'shell' keys, while the bundled schema documents 'provider' (object map with '{env:...}' apiKey options), 'permission' with 'edit'/'bash'/'webfetch' keys, and 'model': 'provider/model-id'. A user copy-pasting the inline examples produces config the skill's own reference says is invalid — concrete guidance with a substantive correctness gap, not the 'minor gaps' of score 4.

Workflow Clarity

The six-step Workflow has an explicit sequence with two validation checkpoints ('jq . opencode.json >/dev/null', smoke test 'opencode run "test"') and a stated error-recovery loop ('If errors surface, re-read config precedence before widening permissions'). This matches the top anchor — clear sequence, explicit validation steps, feedback loop for error recovery — and the destructive/batch cap of 3 does not apply because verification is present.

Progressive Disclosure

Structure is good: a references table with a 'When to Use' column clearly signals three one-level-deep reference files that all exist, and deep field material is correctly kept in references/config-schema.md rather than inlined. It falls short of score 5 because the 'Eval Scenarios' section links to evals/scenario-0..2/task.md and no evals/ directory exists in the bundle — three dead navigation links — and the 'Information Architecture' hub adds a redundant navigation layer.

Specificity

The description lists multiple concrete capability areas — 'deterministic provider setup, model selection, permission policies, formatter behavior, and environment variable handling' — anchored to two named files (opencode.json, AGENTS.md). Coverage of the configuration domain is comprehensive, matching the 'multiple specific concrete actions' anchor; not score 4 because no meaningful capability area is missing, and not below because every action named is concrete rather than generic.

Completeness

Both questions are answered explicitly: the 'what' ('Configure OpenCode via opencode.json and AGENTS.md with...') and the 'when' ('use when editing opencode configuration, setting model/provider defaults, tightening agent permissions, or troubleshooting OpenCode config behavior'). The explicit 'use when' clause with four concrete triggers matches the top anchor exactly, so the missing-'when' cap of 3 does not apply and score 4's 'when could be more explicit' does not fit.

Trigger Term Quality

Natural trigger phrases users would actually say are present: 'editing opencode configuration', 'setting model/provider defaults', 'tightening agent permissions', 'troubleshooting OpenCode config behavior', plus the key filenames opencode.json and AGENTS.md. This covers the domain's natural vocabulary including file names, matching the comprehensive-coverage anchor; it does not fit score 4 because no common phrasing for this domain (config edit, defaults, permissions, troubleshooting) is absent.

Distinctiveness Conflict Risk

The description names a specific tool (OpenCode) and its specific config files, giving it a clear niche with distinct triggers. Overlap risk with general config or Claude Code configuration skills is minimal since every trigger is qualified by OpenCode; not score 4 because there is no meaningful overlap with a closely related skill's trigger set.