inquiry-init

name	inquiry-init
description	Bootstrap a new inquiry-instance in one command — creates the directory tree and composes SCHEMA.md from the memex base plus the inquiry addendum. Assumes memex and big_questions skills are installed user-scoped.

inquiry-init

Purpose

One-command bootstrap for an inquiry-instrument instance. Invoke this skill when the user runs /inquiry-init <path> or otherwise asks to "bootstrap," "initialize," or "set up" a new inquiry (a collection of ~12 big questions). The helper shells out to memex's init_wiki.py for the generic wiki tree, then appends the inquiry schema addendum to SCHEMA.md. That's it — no skill symlinking.

This skill is domain-specific to the inquiry-instrument pattern — questions, contributions, and gap reports. The parallel generic bootstrap is /wiki-init; the helper this skill wraps calls it internally with --pages-dir questions.

Preconditions

• memex is installed user-scoped — specifically, <memex-path>/helpers/init_wiki.py must exist and be executable. The default install location is ~/Development/memex/; the user may pass a different path via --memex-path. The helper checks this first and exits with a clear error if memex isn't found (so no file operations happen on the target directory). If the user sees that error, the fix is to install memex per the memex README (clone + symlink harness skills into ~/.claude/skills/).
• big_questions is installed user-scoped — if this skill is invocable at all (i.e., the user can type /inquiry-init), this precondition is already satisfied implicitly. The big_questions README's install section tells the user to clone the repo and symlink inquiry-elicit, inquiry-gap, and inquiry-init into ~/.claude/skills/. Same pattern as memex.
• The user has write access to the target path. The target may or may not already exist; if it exists, it must be a directory (not a file).
• Python 3 is available on PATH as python3.

Steps

1. Parse the target path

Extract the target path from the user's invocation. Typical shapes:

• /inquiry-init ~/vaults/MyVault/01_Projects/my_big_questions
• /inquiry-init /tmp/scratch_inquiry
• /inquiry-init (path omitted — ask the user)

If the path is missing or ambiguous, ask: "What path? e.g. ~/vaults/MyVault/01_Projects/my_big_questions." Do not guess or default silently to anything potentially destructive.

Expand ~ to the user's home directory before passing to the helper (the helper does not expand shell metacharacters itself).

2. (Optional) --memex-path override

If the user has memex installed somewhere non-standard (not ~/Development/memex/), accept --memex-path <path> and pass it through. Most users will not need this — the default is correct for the standard install pattern documented in the big_questions README.

3. (Optional) --force reseed

If the user passes --force — or explicitly asks to "reseed," "reset," or "refresh" the schema after a schema template update — pass --force through.

What --force does: re-copies SCHEMA.md from the templates (memex base + inquiry addendum). It preserves your questions under wiki/questions/, log.md, index.md, and anything in raw/ or meta/. Use it when you've pulled a schema update and want the instance to pick it up. The final SCHEMA.md will contain exactly one copy of the addendum — not double-appended — so --force is safe on an already-initialized instance.

4. Invoke the helper

Shell out via the bash tool. Always invoke with python3 explicitly (per Decision 5 in the architecture decisions — avoid ambiguous python on macOS where it may resolve to Python 2 or be missing):

python3 <path-to-big_questions>/helpers/init_inquiry.py --path <target> [--memex-path <path>] [--force]

Resolving the helper path. This skill is installed as a symlink into ~/.claude/skills/inquiry-init/ (user-scoped per the big_questions README). helpers/ is not adjacent to the running SKILL.md in the skills directory. Resolve the helper path in this order (same pattern as wiki-init and wiki-ingest):

1. If ./big_questions/helpers/init_inquiry.py exists relative to the current working directory (the user has vendored or symlinked the repo into their instance), use that.
2. Otherwise, resolve the SKILL.md's real path (follow the symlink) and use <big_questions-repo>/helpers/init_inquiry.py — the helpers live as siblings of skills/ in the big_questions repo. On a typical install this is ~/Development/big_questions/helpers/init_inquiry.py.
3. If neither resolves, ask the user where the big_questions repo is installed rather than guessing.

Example — on a typical install bootstrapping a vault-local instance:

python3 ~/Development/big_questions/helpers/init_inquiry.py --path ~/vaults/MyVault/01_Projects/my_big_questions

5. On success — surface the helper's output and suggest next steps

The helper prints a clean structured summary to stdout: a banner line (Inquiry instance initialized at <path>), an [init_wiki output] block relaying the generic wiki helper's own summary, and an [inquiry layer] block listing the addendum status. Relay it to the user verbatim — it already communicates exactly what changed on disk.

Then suggest the natural next step: cd <target> and then run /inquiry-elicit to start the question-elicitation interview. That slash command is available because inquiry-elicit is installed user-scoped (per the big_questions README) — the new instance directory doesn't need its own copy.

6. On failure — surface the helper's stderr

On non-zero exit, the helper writes one or more clear lines to stderr, each prefixed with init_inquiry:. Relay them to the user verbatim — the prefix makes it easy to recognize as coming from this helper rather than the shell, Python, or a nested subprocess.

• If the error says memex harness not found at <path>: gently suggest installing memex. The memex README has the install recipe (clone ~/Development/memex/, symlink the harness skills into ~/.claude/skills/). If the user's memex install is elsewhere, suggest re-running with --memex-path <their-path>.
• If the helper reports that init_wiki itself failed, the nested error will appear in stderr after the init_inquiry: line. Surface both and let the user decide how to proceed; don't retry automatically.
• Any other stderr line — surface it and ask the user how they'd like to proceed.

Idempotency

Safe to re-run this command. If the instance is already initialized, the helper detects it and does nothing destructive — you do not need --force for a plain re-run. Specifically:

• The helper sees that <target>/SCHEMA.md already exists and skips calling init_wiki (which would otherwise hard-fail). It emits a note saying the sub-call was skipped.
• The addendum append checks for the addendum's opening marker in SCHEMA.md. If it's already present, append is skipped.

Use --force only when you explicitly want to reseed SCHEMA.md from the templates (e.g., after a schema-template update). With --force, the final SCHEMA.md still contains exactly one copy of the addendum.

Failure modes

• Target path not provided. Ask the user for an absolute path. Do not assume a default.
• Helper not findable. Follow the three-step resolution order under Step 4. If all three fail, ask: "Where is the big_questions repo installed? I need the path to helpers/init_inquiry.py."
• memex not installed. The helper's precondition check writes init_inquiry: memex harness not found at <path> and exits before touching the target. Surface the error and suggest installing memex (clone + symlink per the memex README) or re-running with --memex-path <path> pointing at an existing install.
• Target exists but isn't a directory. The helper (via init_wiki) surfaces the error. Echo it. The user must rename, move, or pick a different path; do not delete anything on their behalf.
• Write permission denied. Surface the helper's OSError message and suggest the user pick a writable target or adjust permissions.