Lists multiple specific concrete actions: 'triaging assertion failures from mock-coupling failures from genuine regressions' and 'using Python's introspection to automate where safe'. Describes concrete decision-making (update vs. fix vs. delete).

Clearly answers both what ('Updates broken pytest tests...triaging...using Python's introspection') AND when ('Use when a refactor or API change leaves a pile of failing tests and you need to decide update vs. fix vs. delete').

Includes natural keywords users would say: 'pytest', 'tests', 'refactor', 'API change', 'failing tests', 'assertion failures', 'mock'. These are terms developers naturally use when dealing with broken tests.

Highly specific niche: pytest test repair after intentional code changes. The combination of 'pytest', 'refactor/API change', and 'triaging failure types' creates a distinct trigger profile unlikely to conflict with general testing or coding skills.

Extremely lean and efficient. No explanation of what pytest is or how Python works. Every section delivers actionable information Claude wouldn't already know. The cross-reference to java-test-updater avoids repetition.

Provides executable code examples throughout (conftest.py shim, pytest commands, concrete test rewrites). The failure taxonomy table maps specific error messages to specific actions. Copy-paste ready examples for mock decoupling and assertion fixes.

Clear multi-step workflows with explicit validation: snapshot update workflow has numbered steps with review-before-update checkpoint. The 'Do not' section provides explicit guardrails. Output format creates a structured triage checklist with classification steps.

Content is well-organized with clear sections and tables, but everything is inline in one file. The reference to java-test-updater is good, but the skill could benefit from splitting detailed examples (like the conftest shim patterns) into a separate reference file for complex migrations.