# Upgrade report: seriousdevelopment applied (Mode B, scoped run) - Date: 2026-11-08 - Applied by: llm:muse-wonder-1.0 (session floral-aphelion) - Scope: narrowed on user request to AGENTS.md plus version discipline. Audit ran first (read-only); the rest of Mode B is deliberately left out and listed under Follow-ups. ## Grades (A-F, before -> after) | Axis | Before | After | |------|--------|-------| | Spec completeness | D | D (scoped out, follow-up) | | Decision traceability | C | B (CHANGELOG.md, AGENTS.md rules with enforcers) | | Dev tracking | D | D (BUILD-TRACK.md scoped out, follow-up) | | Verification | B | A (311 unittest tests green, Ruff gate clean, both named in AGENTS.md definition of done) | | Doc freshness | B | B (README.md current; doc headers added to new docs only) | | Estimation honesty | C | C (no estimates practice yet, follow-up) | ## Applied changes 1. seriousdevelopment 1.7.1 installed from https://static.molodetz.nl/seriousdevelopment (user scope, `muse skills install`), templates copied into the project, never linked, per the skill rule. 2. AGENTS.md created from the skill template: session preferences as hard rules with enforcers, plus an automatic definition of done that applies to every future task. 3. Version single-sourced: pyproject.toml is the only literal; blackcat.__version__ resolves it live (pyproject next to the source tree, else installed dist metadata). Bump 0.1.0 to 0.2.0 (minor: TUI harness, yolo, todo tool, sessions, context limits). 4. CHANGELOG.md created from the skill template, ascending table, backfilled 0.1.0 and 0.2.0 rows. ## Conflicts (existing rule kept, method rule NOT applied) | # | Existing rule (kept) | Method preference | Why kept | Location | |---|---------------------|-------------------|----------|----------| | 1 | Never commit or stage unless the user asks in the current session | Commit, deploy, verify in the same session | Explicit standing user instruction outranks the method default | AGENTS.md hard rules | | 2 | Zero-dependency stdlib core, setuptools pyproject, unittest | House stack (FastAPI, dataset ORM, pytest tiers, Makefile, container) | Project is a CLI/library agent harness, not a web app; the house stack targets web services | pyproject.toml, AGENTS.md project facts | | 3 | Original prompt strings byte-identical to upstream | Spec and prompts evolve with the project | Standing user instruction from the rename task; only appended guidance allowed | AGENTS.md hard rules | ## Follow-ups 1. SPEC.md backfilled from the code (skill template), zero open gaps, when the user wants the full Mode B baseline. Owner: retoor. 2. BUILD-TRACK.md plus docs/adr/ (ADR 0001: method adoption) for multi-session tracking. Owner: retoor. 3. Doc version headers backfilled on README.md and design.md on their next content edit. Owner: next editor. 4. Makefile and CI wiring for the two named gates (unittest, ruff). Owner: retoor. See also: [AGENTS.md](AGENTS.md), [CHANGELOG.md](CHANGELOG.md).