Conciseness

The body is lean — tables for strategies and budgets, code blocks for config, no explanation of concepts Claude already knows. Minor trimmable filler remains ('The main session stays clean while subagents handle the volume') and subagent delegation appears in both the strategy table and its own section, keeping it below anchor 5.

Actionability

Concrete, executable guidance throughout: exact commands (/context, /compact, /mcp, /resume), a copy-paste settings JSON block, numeric thresholds (>70%, >90%, <10 MCPs, <80 tools), and tool-call budgets (20–80). Coverage spans diagnosis, immediate action, configuration, and recovery.

Workflow Clarity

'Quick Diagnosis' is a numbered sequence with explicit percentage checkpoints, and 'When Context Is Degraded' gives an ordered escalation (compact → new session → config changes). Not 5 because the middle sections are parallel strategy lists rather than a sequenced workflow, and the degraded-context recovery has no re-check step to confirm the fix worked.

Progressive Disclosure

No bundle files exist, and the ~110-line body is organized into clearly labeled sections with a diagnostic entry point up top. Everything is appropriately inlined for a skill this size; minor gaps (behavioral token-efficiency rules could live in a reference file if the skill grows) keep it below a clear-overview-with-signaled-references score of 5.

Specificity

The description states 'Optimize token usage and context management', which names the domain but relies on generic verbs — no concrete actions (compact, audit MCPs, delegate to subagents, trim CLAUDE.md) appear. It sits below anchor 3 ('names domain and 1-2 concrete actions') because no specific capability is enumerated.

Completeness

Both halves are present: a 'what' ('Optimize token usage and context management') and an explicit 'when' clause with three triggers. Not 5 because the 'what' is generic rather than concrete, and not 3 because the 'when' is explicitly stated rather than merely implied.

Trigger Term Quality

'Use when sessions feel slow, context is degraded, or you're running out of budget' contains natural phrases a user would say. Not 5 because common variations like 'context window is full', 'compact', 'token limit', or 'running out of context' are missing.

Distinctiveness Conflict Risk

The context/token-budget niche is fairly distinct, and triggers tied to session slowness and degraded context are unlikely to fire for unrelated skills. Minor overlap risk with general performance or prompt-engineering skills keeps it below 5.