The repo is maintained by agents from multiple vendors, so agent
instructions move out of vendor-specific locations into shared,
git-tracked markdown:
- AGENTS.md is the new entry point for any harness; CLAUDE.md becomes a
symlink to it (Claude Code keeps working unchanged).
- Skill bodies relocate to agent/: VERIFICATION.md (the enforcement
rules, ex .claude/skills/nomarchy), DELEGATION.md (capability tiers
light/standard/frontier, scout/runner role contracts, token economy,
parallel fan-out — consolidates the CLAUDE.md model table, LOOP.md's
economy section, and skill §6.5 into one place; vendor model names
survive only in the per-harness mapping table), THEME-DESIGN.md
(ex .claude/skills/theme, which previously lacked frontmatter).
- .claude/ shrinks to a thin Claude Code adapter: settings, subagent
defs, and skill shims that route into agent/.
- Maps updated: agent/README.md (instructions vs state vs adapters),
docs/README.md, README.md layout tree.
Cleanup: stale old_distro .gitignore entry dropped (dir long gone);
local result*/__pycache__ artifacts removed; JOURNAL.md rotated (29
recent entries kept, 120 older moved to agent/JOURNAL-ARCHIVE.md,
rotation rule documented in the header).
Verification: V0 (docs/meta only) — nix flake check --no-build exit 0
with the tracked symlink; grep sweep confirms no vendor model names
outside DELEGATION.md's mapping table.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Add docs/VISION.md (product themes, v1.0 bar, agent slice hints) and
README maps under docs/ and agent/. Point ROADMAP, BACKLOG, LOOP,
CLAUDE, HARDWARE, and the root README at the split: VISION = intent,
BACKLOG = executable queue, ROADMAP = design history, .claude = harness
only. Mark v1.0 track in BACKLOG PROPOSED for human triage.
Plan/reason/review on the strong model (Fable 5); delegate mechanical,
fully-specified subtasks (grep sweeps, table reconciliation, repeated
edits, log summarization) to cheaper models (haiku/sonnet) via subagent
model overrides — the strong model writes the spec and owns the commit.
Plus narrow-reading and terse-writing rules, and a cheaper --model
allowance for low-stakes headless iterations.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A git-tracked agent/ directory so AI agents can iterate on the distro
unattended (runner-agnostic: /loop, headless claude -p, or a fresh
manual session — all state lives in the checkout, per the distro's own
philosophy):
- LOOP.md — the iteration protocol: orient → pick one BACKLOG task →
verify up the V0–V3 ladder → commit+push main → record. Safety rails
(v1 untouchable, no force-push, no surprise lock bumps) and
stop-and-escalate conditions.
- BACKLOG.md — the forward half of docs/ROADMAP.md reworked into a
prioritized queue (5 NOW / 6 NEXT / LATER / PROPOSED / Decisions);
ROADMAP.md stays the design/decision record + shipped log.
- GOALS.md — the four pillars (stable > reproducible/zero-hidden-state >
effortless config > beautiful), quality bars, non-goals.
- CONVENTIONS.md — coding/design rules (in-flake state, menu placement,
Waybar whole-swap parity, toggle-vs-package, no formatter).
- MEMORY.md — curated hard-won lessons (VM recipes, btrfs-assistant
segfault watch, rofi/WirePlumber/hyprlock gotchas).
- HARDWARE-QUEUE.md — every pending V3 on-hardware check collected from
the ROADMAP, with exact steps, split by machine.
- JOURNAL.md — append-only iteration log, seeded with this bootstrap.
Plus a root CLAUDE.md entry point and README/ROADMAP pointers.
Verified: V0 — docs-only; nix flake check --no-build green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
