Conciseness

The body is efficient: each pattern is a tight When/Pattern/Example block with concrete parameters like "start at 2s, double each attempt" and "Maximum 3 retries", and it avoids explaining concepts Claude already knows. It is not a 5 only because the opening line restates the frontmatter description and a few example lines could be trimmed.

Actionability

As an instruction-only skill it gives concrete, actionable numbered steps with specific thresholds (2s/4s/8s backoff, max 3 retries, max 3 fix-retry cycles) and concrete worked examples ("API call returns 429 → wait 2s → retry..."). It stops short of 5 because some steps stay abstract ("Hand off to the coordinator", no specified log format).

Workflow Clarity

Each pattern is a clearly numbered multi-step sequence, the Diagnose-and-Fix pattern has an explicit validation checkpoint ("Re-run to verify the fix"), the whole skill is built around error-recovery feedback loops with escalation thresholds, and the selection-guide table acts as a checklist — matching the score-5 anchor of explicit validation, feedback loops, and a checklist.

Progressive Disclosure

The skill is well-organized into five numbered pattern sections plus an "Applying These Patterns" section with a selection table, giving easy navigation. It is a self-contained ~90-line single file with no external references; the structure is good but it does not use one-level-deep reference files, and being over 50 lines it falls just outside the simple-skill exception that would warrant a 5.

Specificity

It names the domain ("Standard recovery patterns for all squad agents") but the only action stated is the vague imperative "adapt — don't just report the failure"; no concrete capabilities (retry, fallback, escalate, diagnose) are listed. This matches score 2 (names domain, actions minimal/generic) and falls short of 3, which requires 1-2 concrete actions.

Completeness

It answers both "what" ("Standard recovery patterns for all squad agents") and "when" ("When something fails") with an explicit trigger clause, so it clears the missing-trigger cap. It is a 4 rather than 5 because the "when" is general and does not enumerate concrete trigger scenarios (API failures, build breaks, missing tools).

Trigger Term Quality

The description offers only a couple of generic keywords ("recovery patterns", "fails", "failure") and omits the natural phrases a user would actually say ("retry", "error handling", "fallback", "something went wrong"). It sits at score 2 rather than 3 because the keyword set is thin and lacks common variations.

Distinctiveness Conflict Risk

Scoping to "squad agents" gives it a clear niche and mostly distinguishes it from other skills. The trigger "When something fails" is broad and could overlap with related reliability skills, leaving minor conflict risk rather than the minimal risk of a 5.