[The Framework] Bootstrap: wrap the user prompt instead of a system-prompt preamble

[suleimansh]

Part of #297 (Bootstrap mode).

Rom's steer on #297: treat Claude Code as a black box and wrap the user prompt, rather than injecting via --append-system-prompt / --system-prompt. Instructions carried in the prompt itself outweigh "contextual" system-prompt text the agent is tempted to skip.

#448 shipped bootstrap as a forceful system-prompt preamble (BOOTSTRAP_PREAMBLE prepended in systemPromptBlock). This reworks the mechanism to Rom's approach:

• New wrapBootstrapPrompt(userPrompt): rewords the intent into instructions (analyze the request, showMarkdown() the analysis + plan, then await approval; no code before approval). Original request rides at the end unchanged.
• Direct-prompt path (runPrompt): wrap the first user prompt when bootstrap is on.
• Build path (runFramework): a short JSON-compatible bootstrap note on the architect turn; the existing plan-approval gate already stops for the plan.
• Drop the system-prompt preamble path (BOOTSTRAP_PREAMBLE, the bootstrap branch of systemPromptBlock).
• CLI --bootstrap flag + daemon flag mapping unchanged.

Verify: a real claude -p run in an empty dir with the wrapped prompt stops for a plan (no files written).

[brillout]

FYI I'm trying to design a single system prompt to be always injected, not only for bootstraping.

Preview:

ADD_ANALYZE_ENTRY: Add entry to the ANALYZE_PROMPT.md table

Analyze the user prompt, follow the instructions, create ANALYZE_PROMPT.md that is a list of YES/NO entries and <SHOW_MD>

## Ambiguous prompt

If it isn't clear what you should do (e.g. unclear scope, unclear user prompt), make a list of interpretations sorted by plausibility, <SHOW_CHOICES>, <AWAIT>

<ADD_ANALYZE_ENTRY> whether YES/NO the prompt is ambiguous

## Large scope

- If the scope of what you'll work on is *large*, create a `PLAN_<SESSION_NAME>.agent.md` of what you'll work on, <SHOW_MD>, <AWAIT>
- If the scope is potentially *very large* (e.g. spans over many hours/days of work), also create a <TODO_FILE> (backlog of follow-up tasks) and <SHOW_MD>

<ADD_ANALYZE_ENTRY> whether the work scope is small, large, or very large

[suleimansh]

Makes sense, this subsumes bootstrap: an empty dir is just 'scope = very large' -> PLAN + await from the always-on analyze. Happy to hold #458 (the wrap-the-prompt PR) and fold it into the unified prompt instead.

Once you've got the full ANALYZE_PROMPT.md prompt, want me to prototype it against a few real runs (ambiguous / large / empty dir) like we did for the bootstrap wrap? Mostly to see how reliably it fills the YES/NO table.

[brillout]

Yes, that'd be great. Let me finish it (I also have a new idea for it).