## Parent
#1
## Type
AFK
## What to build
Remove or qualify brittle numeric claims and generated-map wording that drift as the repository evolves. This includes skill counts, E2E test counts, and fork/upstream LLM map guidance.
## Acceptance criteria
- [x] Replace broad hard-coded skill counts with scoped wording or source-backed counts.
- [x] Replace hard-coded E2E counts with count-free or script-derived wording.
- [x] Clarify generated llms.txt and llms-full.txt authority and fork-specific LLMS_REPO_BASE regeneration guidance.
- [x] Ensure README, install docs, agent docs, and release/testing docs do not contradict each other on these counts.
## Blocked by
- #2
- #3
- #4
## Implementation status
Implemented in PR #2212.
Evidence:
- `docs/docs-consolidation/08-brittle-counts-and-generated-maps.md` records removed/qualified count claims.
- `README.md`, `AGENTS.md`, `INSTALL_FOR_AGENTS.md`, `CLAUDE.md`, `docs/TESTING.md`, `scripts/llms-config.ts`, and `scripts/build-llms.ts` were updated where needed.
- `llms.txt` and `llms-full.txt` were regenerated from source.
Tracker note: this issue is closed as complete in the fork because the work is present in garrytan/gbrain#2212. Upstream merge/review remains tracked by https://github.com/garrytan/gbrain/pull/2212.
Implemented in commit aeabfe67 (`docs: qualify generated map and count claims`).
What changed:
- Replaced broad count claims in README, CLAUDE.md, docs/TESTING.md, and generated LLM map source.
- Clarified that `llms.txt` and `llms-full.txt` are generated from `scripts/llms-config.ts` and that forks should regenerate with `LLMS_REPO_BASE`.
- Added `docs/docs-consolidation/08-brittle-counts-and-generated-maps.md` as the #13 implementation note.
- Regenerated `llms.txt` and `llms-full.txt`.
Validation:
- `bun run build:llms` passed and rewrote both maps.
- `git diff --cached --check` passed before commit.
- Targeted `rg` check found no stale current-doc phrases: `29 skills`, `43 skills`, `26 fat`, `3650+ tests`, `30+ tools`, `~47 operations`, `306 skills, ~21K`, `300+ skills`.
- Staged-diff secret-pattern scan returned no hits.
Test note:
- `bun test test/build-llms.test.ts` could not run in this checkout because Bun cannot resolve package `ai` from `src/core/ai/gateway.ts`; no dependency install was performed.
Marked complete for the docs-only branch in garrytan/gbrain#2212. Acceptance criteria are checked in the issue body with evidence and proof limits. Upstream review/merge remains tracked at https://github.com/garrytan/gbrain/pull/2212.