fable-status

Show what oh-my-fable is doing right now, in one table, without changing anything. Plugin version, where the rules live, this session's mode and why, effort and where it comes from, whether the rules

SKILL.md

GitHub

fable-status · what is in effect (read-only)

This skill writes nothing. It answers "is it working, and how?" in one table so the user never has to inspect files or hook output by hand. Do not run /fable-setup from here; only point to it when a row says so.

Step 1 · One batch of reads

  • bash "${CLAUDE_PLUGIN_ROOT}/hooks/session-start.sh" --status, exactly as written, from the project root (the hook reads the current folder as the project; written this way it runs without a permission prompt). It computes everything: version, config dir, entrypoint, mode, delivery, rules file version, effort and its source, subagent delivery, notices.
  • claude --version.
  • The frontmatter of ./.claude/agents/*.md and ~/.claude/agents/*.md (or under $CLAUDE_CONFIG_DIR), only whether each file sets effort:.
  • The CLAUDE.md files the hook checks (~/.claude/CLAUDE.md or $CLAUDE_CONFIG_DIR/CLAUDE.md, ./CLAUDE.md, ./.claude/CLAUDE.md) for conflicting rules: narration suppression ("hold findings for the final response", "no closing recap"), anti-formatting rules ("no bullets", "no headers"), "ask before every step".

Step 2 · One table, in the user's language

Row Value Note when it matters
Plugin version from --status, Claude Code version below 2.1.258: say the hook needs a newer Claude Code
Rules rules_source in plain words: "rules file + hook", "hook only", "CLAUDE.md section", "your own rules file (hook silent)", "disabled" show base_file when there is one
Subagents the subagents line
Mode mode, and when it says (auto) add the reason: entrypoint sdk-* → unattended, none → interactive pinned modes: say "pinned in oh-my-fable.json"
Effort effort and effort_source (live = read from this session, exact; saved = modelSettings for this model; settings = top-level effortLevel; env = CLAUDE_CODE_EFFORT_LEVEL; model default = nothing saved, or a top-level effortLevel this model ignores; X → Y by maxEffortLevel = capped) Opus 5.5 and Sonnet 5.5 default to medium in Claude Code and ignore a top-level effortLevel in user settings; raise per session with /effort high then s. Fable 5.1 and the other models default to high. At xhigh/max the hook adds block N (stop once the checks pass, no self-started review rounds)
Subagents' effort note only: subagents inherit this session's effort unless their agent file sets effort: show when an agent file from Step 1 has no effort:, with the file names
Rules file rules_file_version vs plugin_rules_version lower → "run /fable-setup refresh"; higher → "this session runs an older plugin; start a new session, after updating the plugin if needed (no refresh)"
CLAUDE.md conflicts count, then one line per conflict with the file and the suggested replacement none → "none"
Notice the notice line when it is not (none)

Then close with exactly one line, translated: "Nothing was changed. /fable-setup changes settings, /fable <request> improves a single request."

One status line: DONE (or DONE_WITH_CONCERNS when a row needs action).