word_stats_demo¶
Demo skill showing the python preprocessor step pattern. Produces a short commentary grounded in precomputed text statistics.
Entry¶
review
Final output¶
text_review — commentary (1–3 sentence prose citing the exact precomputed values).
How it composes¶
A single review phase. The stats.py preprocessor (safe mode, 5 s timeout) runs deterministically before the LLM, injecting char_count, word_count, line_count, longest_line_chars, and estimated_tokens into data.stats. The LLM writes commentary citing the exact values — it does not recount anything.
Caveats¶
Safe-mode Python; no special permissions required beyond the default. This skill is a reference example, not a production utility.