podcast-generation

Fix: Add a '## Writing Style' or '## Anti-Slop' section

Fix: Add pass/fail criteria or measurable outcomes (e.g., 'must return valid JSON', 'output < 100 lines')

Fix: Add a criteria section (checklist, rubric, or pass/fail conditions)

Fix: Document the output format (template, schema, or example output)

Fix: Verify the path or remove the stale reference

Fix: Add a '## Limitations' or '## Error Handling' section describing when the skill won't work or how it handles failures

Fix: Add explicit DO and DON'T lists for writing patterns

Fix: Add rollback instructions: how to undo changes if something goes wrong (e.g., git revert, backup/restore steps)

Fix: Add idempotency guidance: state whether running the skill twice produces the same result or causes side effects

Fix: Add a verify/validate step after the main operation (per best practices: run validator, fix, repeat)

Fix: Document what the skill expects as input (format, required fields, constraints)

Fix: Document what the skill produces (format, structure, side effects)

Fix: Clarify whether the skill works standalone, in pipelines, or both

Fix: Add example input/output pairs or <example> blocks to make the skill testable

Fix: Consider gerund-style names like 'processing-pdfs' or 'analyzing-logs' (per best practices)

Fix: Add confidence indicators (e.g., confidence: high/medium/low) so orchestrators can route or retry

Fix: Define a structured output format (JSON, YAML, or table) so downstream tools can parse results