sg-toggle-learn

name	sg-toggle-learn
description	Toggle the sg-learn stage on or off in the super-gsd workflow. Flips super_gsd.skip_learn in .planning/config.json so sg-next / sg-status / the stop hook include or skip the sg-retro retrospective after review. Accepts on|off, or no argument to flip.
argument-hint	[on|off] - 'on' includes learn, 'off' skips it, empty flips the current state.

Detect the user's input language and respond in that language throughout this skill's output. - Korean input → respond in Korean - English input → respond in English - Mixed input → match the dominant language Enable or disable the sg-learn stage in the super-gsd workflow by writing the super_gsd.skip_learn boolean in .planning/config.json. ON means the retrospective runs (skip_learn=false, the default); OFF means sg-learn is skipped (skip_learn=true). With no argument, flip the current value. All other config keys and 2-space formatting are preserved.

<execution_context> Self-contained. Reads and writes .planning/config.json only. Does not touch HANDOFF.md or any other workflow file. </execution_context>

1. **Parse the desired state from $ARGUMENTS.** ```bash ARG=$(printf '%s' "$ARGUMENTS" | tr '[:upper:]' '[:lower:]' | tr -d '[:space:]') case "$ARG" in on|enable|true|include) DESIRED="on" ;; off|disable|false|skip) DESIRED="off" ;; "") DESIRED="toggle" ;; *) echo "Unknown argument '$ARGUMENTS'. Use: on | off | (empty to toggle)."; exit 1 ;; esac ```

2. Read-modify-write config.json (macOS compatible — node only, no jq). The underlying flag is skip_learn (inverted: learn ON ⇒ skip_learn=false). The effective post-implementation sequence is printed for confirmation.

   RESULT=$(DESIRED="$DESIRED" node -e '
   const fs=require("fs"), path=require("path");
   const p=path.join(process.cwd(),".planning","config.json");
   let cfg={};
   try{cfg=JSON.parse(fs.readFileSync(p,"utf-8"));}catch(e){cfg={};}
   cfg.super_gsd=cfg.super_gsd||{};
   const cur=!!cfg.super_gsd.skip_learn;
   const d=process.env.DESIRED;
   let skip; if(d==="on")skip=false; else if(d==="off")skip=true; else skip=!cur;
   cfg.super_gsd.skip_learn=skip;
   fs.mkdirSync(path.dirname(p),{recursive:true});
   fs.writeFileSync(p, JSON.stringify(cfg,null,2)+"\n");
   const tdd=!!cfg.super_gsd.tdd_mode, sr=!!cfg.super_gsd.skip_review, sl=skip;
   const seq=["execute"]; if(tdd)seq.push("tdd"); if(!sr)seq.push("review"); if(!sl)seq.push("learn"); seq.push("ship");
   process.stdout.write((skip?"off":"on")+"\t"+seq.join(" → "));
   ' 2>/dev/null)
   if [ -z "$RESULT" ]; then echo "Failed to update .planning/config.json"; exit 1; fi
   STATE=$(printf '%s' "$RESULT" | cut -f1)
   PATHSEQ=$(printf '%s' "$RESULT" | cut -f2)
   echo "learn:$STATE"
   echo "path:$PATHSEQ"
   

3. Report the new state in the user's language. Surface the result as prose in the user's input language. Keep the command token sg-learn, the flag name skip_learn, and the stage sequence tokens (execute → … → ship) verbatim in English.

   • When STATE=on: tell the user the sg-learn stage is now included in the workflow.
   • When STATE=off: tell the user the sg-learn stage is now skipped.
   • Always show the resulting workflow sequence from path: so the user can confirm.

<success_criteria>

1. With no argument, super_gsd.skip_learn is flipped relative to its current value (absent treated as false).
2. With on/off, the stage is set to included/skipped respectively (skip_learn=false/true), regardless of prior value (idempotent).
3. Existing config.json keys and 2-space indentation are preserved; .planning/config.json is created if absent.
4. The new stage state and the resulting execute → … → ship sequence are reported in the user's language, with command/flag/stage tokens verbatim.
5. HANDOFF.md and all other workflow files are untouched. </success_criteria>