/fable-eval

Run the fable-mode eval suite (probes → pairwise judge → report). Costs tokens — runs headless claude many times. Optional arg: a probe id substring to filter.

BOOST

rennf93@rennf93

GitHub

From plugin

fable-mode

342 skills1 agent2 commands7 hooks

> /plugin marketplace add rennf93/opus-fable-playbook
> /plugin install fable-mode@opus-fable-playbook

Command definition

fable-eval.md

description: Run the fable-mode eval suite (probes → pairwise judge → report). Costs tokens — runs headless claude many times. Optional arg: a probe id substring to filter.

Run the fable-mode convergence evals. `$ARGUMENTS` may contain a probe id filter (e.g. `03`), and `--yes` to skip the cost confirmation.

1. Unless `--yes` was passed, tell the user how many probe runs this is (matched probes × 2 candidate modes + judging) and ask to proceed. 2. Ensure goldens exist in `evals/golden/` for the matched probes; for any missing, generate: `evals/run-probe.sh <probe> golden evals/golden`. 3. Run baseline + fable candidates into `evals/results/<today>/`, then judge each against its golden, per evals/LOOP.md steps 1–2 (respect the filter). 4. Run `evals/report.sh evals/results/<today>/` and present the table. 5. Interpret per LOOP.md step 4: name the weakest dimension and the exact doctrine section/hook/skill to strengthen. Do not apply changes — recommend them.

Read it on GitHub ↗

Stats

Bash C++GitHub Anthropic

Stars

Forks

Active

Maintenance

Shell

Language

MIT

License

8d ago

Last commit

4mo ago

Created

Repo: rennf93/opus-fable-playbook

In this plugin

7 hooks 3 skills 2 commands 1 agent

See everything inside

Ships withfable-mode

Make Claude Opus 4.8 in Claude Code behave as much like Claude Fable 5 as possible. The doctrine was transcribed by Fable 5 itself; hooks enforce it at the harness level; an eval loop measures convergence against golden Fable transcripts.

343 skills1 agent

Get the whole plugin

Stars on GitHub