- 06-tasks: add per-task detail sheets for Phase 0/1 (concrete init
commands, BLOCKED preconditions, verification commands with expected
results) and a constraint/verification quick table for Phase 2/3;
replace subjective acceptance wording with checkable items
- 04-architecture: promote 3.4 to an authoritative filter-param
contract (names, values, AND combination, invalid-value handling);
add required/optional/default field annex with 0-5 validators to
3.3; state AiCodingScore is not a model in MVP; record structure
decision (keep wagtail-start home/search apps, business models in
new core app, no Dockerfile)
- 03-tech-stack: add version-pinning policy row for T-001
- routes/02-requirements: point filter wording back to the 3.4
contract instead of maintaining copies
- T-104: human-input boundary - seed projects, scores and verdicts
must be user-confirmed; BLOCKED without a confirmed list
- current-state snapshot updated; maintenance record appended to
progress.md
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Replace misspelled/ineffective ignore file with a real .gitignore:
deploy/ (contains SSH private key), *.pem, .env*, and Python/Django/
Wagtail runtime artifacts are now excluded from version control
- Add .gitattributes (* text=auto eol=lf) to stop CRLF/LF churn that
made all 26 tracked files show spurious full-file diffs
- Correct repo root path /mnt/d/OPC/skelet -> /mnt/d/opc_project/skelet
in AGENTS.md, 00-ai-start-here, 03-tech-stack, 90-harness-reference,
current-state (progress.md history left as-is, append-only)
- Update current-state.md snapshot and append maintenance record to
progress.md per harness rules
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>