Output Improvement¶
Use primr improve (or --improve) to run a post-generation quality pass on existing .md / .txt outputs.
Commands¶
# Deterministic cleanup + QA metrics
primr improve "output/Company_Strategic_Overview_03-06-2026.md"
# Add an agentic review pass first (find weak sections, then tighten)
primr improve "output/Company_AI_Strategy_AZURE_03-06-2026.md" --improve-agentic --dry-run
# Quote the bounded QA refinement loop
primr refine "Company" --dry-run
# Overwrite the original file instead of writing *_improved
primr improve "output/Company_Strategic_Overview_03-06-2026.md" --in-place
Plain primr improve is deterministic, local, and model-free. The
--improve-agentic and refine paths can call models. Their dry runs quote the
maximum task shape without changing files. Execution repeats the quote,
requires explicit approval, and honors --budget; automation uses
--skip-confirm only after approval. With --json, dry runs, approval-required
refusals, and approved results each emit one versioned JSON object.
What This Does¶
- Removes internal placeholder/source artifacts that should not ship (
Analysis Context,vendor-research,citation inventory, etc.) - Normalizes and validates citations for reports
- Applies strategy consistency checks (including budget-total mismatch detection)
- Runs a deterministic salvage pass before blocking output, so recoverable markdown issues are auto-cleaned
- Applies an artifact shipping gate before DOCX render and validates the rendered DOCX text afterward
- Holds back dirty DOCX files but still saves
.md/.txtplus a validation sidecar when issues remain - Prints deterministic QA summary (
gate=PASS|WARN) before writing output