Work log: Phantasmagoria — March 30, 2026

What shipped today

A full-spectrum scout run identified 30+ findings across error-handling, security, dead code, test coverage, feature gaps, dependency health, and risk management. The session was primarily diagnostic — understanding the codebase’s health after the intensive v2 push — with one concrete fix shipped.

Deep codebase audit. Seven parallel scout agents scanned every dimension listed in .scout.yml. The most significant finding was that resource_gain_scaled — the new default effect type per the March 28 decision — is completely invisible to all three validation layers. The feature-gaps scan also found that balance checks are advisory-only (warnings, never errors), contradicting the “reject weak YAML” decision. The dead-code scan identified an entire orphaned file (validation.py, untouched since December), 4 dead accessor functions in option_roller.py, dead prompt-building functions in context.py, and two dead config blocks in config_mod.yaml. The test-coverage scan revealed that stage2_runner.py (the core generation pipeline), renderability_contract.py (the Stage 2B/3 gate), and followup_generator.py all have zero test coverage. Python 3.9 EOL was flagged as the only high-severity dependency issue.

Error handling fix. phmg/config.py was using sys.exit(1) in library functions, making errors uncatchable by callers like check_stage_contracts.py. Replaced with a proper ConfigError exception class. Tests updated from SystemExit to ConfigError assertions. All 451 tests pass.

Issue pipeline stocked. Five issues filed from scout findings: #261 (resource_gain_scaled validation — needs prep), #262 (sys.exit fix — shipped), #263 (dead code cleanup — ready), #264 (RISKS.md creation — ready), #265 (Python 3.9 upgrade — needs prep).

Completed

• #262 — Replace sys.exit(1) with exceptions in phmg/config.py

Carry-over

• Stage 3/4 split plan still unexecuted (from March 29 carry-over)
• #263 — Dead code cleanup (ready-for-dev, next in queue)
• #264 — Create RISKS.md risk register (ready-for-dev)
• #261 — resource_gain_scaled validation (needs prep, high impact)
• #265 — Python 3.9 EOL upgrade (needs prep)

Risks

• resource_gain_scaled validation hole — the most commonly generated effect type bypasses all three validation layers. Filed as #261.
• Stage 2 pipeline has zero test coverage — stage2_runner.py, renderability_contract.py, and followup_generator.py are all untested. A bug in any of these ships broken events silently.
• Python 3.9 is EOL — no security patches since October 2025. Dependencies could drop support at any time.

Flags and watch-outs

• The test-coverage scan found 10+ modules with zero test files, including critical pipeline components. The 451-test count looks healthy but coverage is concentrated in the renderer/validator layers; the generation pipeline is largely untested.
• parse_json_response() is duplicated in 3 places with potentially different edge-case behavior.
• The celestial_equinox backup files are still untracked in git status.

Next session

1. #263 — Dead code cleanup — quick win, clears 7 dead items, already spec’d and ready-for-dev.
2. #264 — Create RISKS.md — consolidate the 6+ scattered risks into a formal register.
3. #261 — resource_gain_scaled validation — prep and implement. This is the highest-impact finding: the new default effect type has zero validation.
4. Stage 3/4 split — still carrying over from March 29. The plan exists at .claude/plans/shimmering-skipping-oasis.md.
5. Consider test coverage — stage2_runner.py and renderability_contract.py are the two most critical untested modules.