11. Stack multiple skills

⋯ ---name: foxdescription: Reports the den location for the project's fox, demonstrating skill stacking.--- Include the word "den" in your answer, on its own line. Do not call any tools.

⋯ ---name: walkdescription: Takes an animal for a walk, demonstrating shared arguments in a skill stack.--- Include this sentence in your answer, on its own line: Take the $0 for a walk. Do not call any tools.

⋯ """Stage 9: slash commands, level 2 (invoke)Stage 11: stacking multiple skills in one prompt A "/name" prompt loads that skill's SKILL.md body instead of the rawprompt — see skills.py for how resolution and progressive disclosure work."/fox /walk 4127" now expands into one user message per skill — seeskills.py for how the expansion and argument-sharing rules work.""" import argparse⋯ def main():     client = OpenAI(api_key=API_KEY, base_url=BASE_URL)     messages = [{"role": "user", "content": args.prompt}]     skill_body = resolve_slash_command(args.prompt)    if skill_body is not None:        print(f"[agent] resolved slash command to skill body ({len(skill_body)} chars)", file=sys.stderr)        messages[0]["content"] = skill_body    skill_bodies = resolve_slash_command(args.prompt)    if skill_bodies is not None:        print(f"[agent] resolved {len(skill_bodies)} skill invocation(s)", file=sys.stderr)        messages = [{"role": "user", "content": body} for body in skill_bodies]    else:        messages = [{"role": "user", "content": args.prompt}]     skills = discover_skills()    system_prompt = build_skills_system_prompt(skills)

⋯ """Stage 10: skill arguments ($ARGUMENTS, $0, $1, ...) A slash command can pass arguments after the skill name:"/feed rabbit carrots" invokes "feed" with args ["rabbit", "carrots"].The body can reference them with placeholders, substituted before thebody is sent to the model:   $ARGUMENTS         everything after the skill name, joined by spaces  $ARGUMENTS[0], [1] a single argument, by position  $0, $1, ...        shorthand for the same thing Order matters when substituting: $ARGUMENTS[0] must be handled before thebare $ARGUMENTS, or a naive string replace would mangle it into"rabbit carrots[0]".Stage 11: stacking multiple skills in one prompt A prompt can invoke more than one skill at once: "/rabbit /fox 4127"loads both bodies, and the trailing "4127" reaches each of them as$ARGUMENTS. Expansion runs left to right from the start of the prompt —every token that names a real skill gets expanded; the first token thatdoesn't name a skill ends the run, and it (plus everything after it)becomes the shared argument text for every skill that WAS expanded.   "/rabbit /fox 4127"        -> expands [rabbit, fox], args = "4127"  "/rabbit 4127 /fox"        -> expands [rabbit],       args = "4127 /fox"  "/rabbit /owl 4127" (no owl skill) -> expands [rabbit], args = "/owl 4127"""" import os⋯ def load_skill_body(name, skills_dir=SKILLS_DIR):  def resolve_slash_command(prompt, skills_dir=SKILLS_DIR):    """If prompt starts with '/', resolve the first word as a skill name,    substitute any argument placeholders in its body, and return that as    the user message content.    """If prompt starts with one or more "/skill-name" tokens, expand them    all and return a list of substituted bodies, one per skill — in the    order they appeared, each with the same trailing argument text.     Returns None if the prompt isn't a slash command or the named skill    can't be found, so the caller falls back to the raw prompt.    """    if not prompt.startswith("/"):        return None    Expansion stops at the first token that isn't a real skill name; that    token and everything after it becomes the shared $ARGUMENTS text.     rest = prompt[1:].strip()    if not rest:    Returns None if the prompt doesn't start with a recognized skill    invocation at all, so the caller falls back to the raw prompt.    """    tokens = prompt.split()     expanded_names = []    i = 0    while i < len(tokens) and tokens[i].startswith("/"):        candidate = tokens[i][1:]        if load_skill_body(candidate, skills_dir) is None:            break        expanded_names.append(candidate)        i += 1     if not expanded_names:        return None     parts = rest.split()    skill_name, args = parts[0], parts[1:]     body = load_skill_body(skill_name, skills_dir)    if body is None:        return None    args = tokens[i:]     return substitute_arguments(body, args)    return [substitute_arguments(load_skill_body(name, skills_dir), args) for name in expanded_names]  def substitute_arguments(body, args):