Conciseness

The body is dense with harness-specific, non-inferable facts and assumes competence throughout, but a few passages restate rules or run discursive — e.g., 'Reading a shipped composition is the intended way to start; writing to one is not, and neither is editing the host composition to work around a preset limitation' repeats the section's own prohibition — leaving minor instances that could be trimmed.

Actionability

Fully executable guidance: a complete copy-paste JS plugin for mount-validation, exact API calls ('cordis_inspect what:"api" name:"agentPresets"', 'standingKeyFor(id)', 'copy(from, id, name?)'), ready-to-copy YAML templates for the subagent rows, and the literal error-message strings to expect — covering the common cases.

Workflow Clarity

The numbered authoring sequence (copy, expect sandbox, write metadata, edit rows, mount-validate) is explicitly capped by the 'standingKeyFor(id) is the check' validation step, with a four-failure-mode catalog and feedback loops ('read the rejection — it names the offending service') plus a post-validation handoff to a real session.

Progressive Disclosure

Well-ordered, clearly headed sections make the ~150-line single file easy to navigate with no nested references, but it is a monolith — the roster API details, the error-message catalog, and the disabled subagent templates are candidates for one-level-deep reference files that would keep SKILL.md a leaner overview.

Specificity

The description enumerates six concrete actions — 'creating, changing, or validating a Cordis composition', 'writing or editing an agent preset', 'adding or removing a plugin row', 'deciding whether something belongs to the host composition or to one session', 'checking whether a preset you authored actually mounts', and 'diagnosing a row that mounted but contributed nothing' — comprehensively covering authoring, editing, validation, and diagnosis.

Completeness

It explicitly answers 'when' with a detailed 'Use when...' clause and 'what' through the enumerated concrete actions, matching the anchor for both halves with concrete trigger phrases.

Trigger Term Quality

Natural user phrasings like 'agent preset', 'plugin row', 'host composition', and 'a row that mounted but contributed nothing' are present with verb variations (create/change, write/edit, add/remove), but file-level terms a user might naturally mention such as 'cordis.yml' or 'preset.yml' are absent, leaving a few natural terms missing.

Distinctiveness Conflict Risk

'Cordis composition for this harness' carves out a clear niche with domain-specific triggers (presets, plugin rows, mount validation) that no generic skill would compete for, giving minimal conflict risk.