BACKLOG.md

Backlog

Prioritized list of improvements for autonomous agents to pick up. Check gh issue list --label agent-ready first — GitHub Issues take priority over this list.

Mark items as done with strikethrough when completed via PR.

Priority 1 (High Impact)

1. Increase test coverage to 80% — identify untested branches in src/agents/ and src/orchestrator.py, add unit tests following existing mock patterns in tests/conftest.py
2. Add --dry-run CLI flag — show what the pipeline would do without calling OpenAI or writing files; useful for validating config and arguments
3. Add retry logic to WriterAgent — ResearchAgent uses tenacity.retry for web calls; WriterAgent has no retry on LLM failures, add the same pattern

Priority 2 (Medium Impact)

1. Add input validation for CLI arguments — validate topic length (non-empty, reasonable max), validate style values, surface clear error messages
2. Add health CLI command — verify API keys work (OpenAI, Unsplash, Medium) with lightweight API calls before running the full pipeline
3. Consolidate setup.py into pyproject.toml — move package metadata, dependencies, and entry points from setup.py to pyproject.toml (PEP 621)

Priority 3 (Incremental)

1. Pin dependency versions — replace >= with ~= or exact pins in requirements.txt for reproducible builds
2. Extract prompt templates — move inline prompt strings from agents into src/prompts/ directory for easier tuning without touching agent logic
3. Add structured JSON logging option — add --log-format json flag for machine-readable log output
4. Add type stubs / py.typed marker — enable downstream type checking for anyone importing the package