Name Claim

name-claim

The name to claim is: $ARGUMENTS

The sibling template directories are:

• npm: $SKILL_DIR/../npm-claim/contents/
• pypi: $SKILL_DIR/../pypi-claim/contents/
• crate: $SKILL_DIR/../crate-claim/contents/

Contract

This skill scaffolds, the user publishes. The orchestrator does NOT run the publish scripts. Publishing to npm, PyPI, or crates.io requires interactive auth (browser OAuth, terminal prompts, 2FA codes) that needs a real TTY. The Bash tool has no TTY, so any auto-publish attempt would hang or fail.

The skill's job ends after step 5: print three copy-pasteable commands and stop. The user runs them in their own terminal as one paste.

Steps

1. Validate and compute substitution values

If $ARGUMENTS is empty or whitespace, tell the user to provide a name and stop.

Compute three values, used throughout:

• NAME = $ARGUMENTS (kebab-case as given)
• TITLE = title case (e.g. identity-matters becomes Identity Matters)
• MODULE = NAME with hyphens replaced by underscores (Python module name)

2. Check all three registries in parallel

Run these three Bash calls in a single message (parallel):

# npm
npm view "$ARGUMENTS" version 2>/dev/null; echo "EXIT:$?"

# PyPI
curl -s -o /dev/null -w "%{http_code}" "https://pypi.org/pypi/$ARGUMENTS/json"

# crates.io
curl -s -o /dev/null -w "%{http_code}" -A "helioy-name-claim (https://github.com/littleorgans)" "https://crates.io/api/v1/crates/$ARGUMENTS"

The crates.io -A User-Agent flag is mandatory. crates.io returns 403 for any request without a User-Agent header, so a flag-less call would misreport an available name as unknown/taken. Keep -A on every crates.io call here and in step 2b.

Interpret each:

If crates.io returns 403, the User-Agent header was dropped: re-run the command with the -A flag exactly as shown rather than recording a verdict. Any other status: report as unknown and treat as taken for this run (do not scaffold an unknown registry).

2b. Owner lookup for taken registries

For any registry that came back taken, run its owner query so the user can immediately tell whether the existing claim is theirs. Batch all taken-registry lookups in a single parallel Bash message:

npm:

npm view "$ARGUMENTS" maintainers 2>/dev/null

pypi:

curl -s "https://pypi.org/pypi/$ARGUMENTS/json" | python3 -c "import sys,json; d=json.load(sys.stdin)['info']; print(d.get('author') or d.get('maintainer') or 'unknown')"

crate:

curl -s -A "helioy-name-claim (https://github.com/littleorgans)" "https://crates.io/api/v1/crates/$ARGUMENTS/owners" | python3 -c "import sys,json; d=json.load(sys.stdin); print(', '.join(u['login'] for u in d['users']))"

Skip this step entirely if all three registries are available.

3. Print availability

Print:

$ARGUMENTS
  npm        <available | taken (owner: <name>) | unknown>
  pypi       <available | taken (owner: <name>) | unknown>
  crates.io  <available | taken (owner: <name>) | unknown>

If an owner matches the user (e.g. srobinson on crates.io, matching npm maintainer, or matching PyPI author email), append , that's you inside the owner parentheses so the user knows the existing claim is theirs. Example: taken (owner: srobinson, that's you).

If no registry is available, say so plainly and stop.

Otherwise print:

Claiming on: <comma-separated list of available registries>
Crates.io is permanent (yank only, no unpublish).

4. Scaffold all available registries

Create one shared directory and fill it with the per-registry artifacts. The file sets do not collide; only the publish script is suffixed by registry.

Make the parent directory first:

mkdir -p ~/.name-claim/$ARGUMENTS/src

Then, for each available registry, run the matching block. Substitute the literal values for NAME, TITLE, and MODULE into the sed commands before running.

Important sister-script edit. The sibling publish.sh templates each end with rm -rf /tmp/claim-{{NAME}}. Fine for a single-registry claim, fatal here: the first script to finish would delete the shared dir mid-flight while the user is still running the other two. Strip that line during scaffolding (it stays in the sister skills, which run solo). The user can delete ~/.name-claim/$ARGUMENTS/ by hand once all three publishes are done.

The npm login line in the npm publish script is kept: the user runs the script in their own terminal, where the browser/2FA flow works fine.

npm (if available):

sed -e 's/{{NAME}}/NAME/g' -e 's/{{TITLE}}/TITLE/g' \
  $SKILL_DIR/../npm-claim/contents/package.json \
  > ~/.name-claim/$ARGUMENTS/package.json
sed 's/{{NAME}}/NAME/g' \
  $SKILL_DIR/../npm-claim/contents/placeholder.js \
  > ~/.name-claim/$ARGUMENTS/$ARGUMENTS.js
sed -e 's/{{NAME}}/NAME/g' -e '/^rm -rf /d' \
  $SKILL_DIR/../npm-claim/contents/publish.sh \
  > ~/.name-claim/$ARGUMENTS/publish-npm.sh
chmod +x ~/.name-claim/$ARGUMENTS/publish-npm.sh

pypi (if available):

sed -e 's/{{NAME}}/NAME/g' -e 's/{{TITLE}}/TITLE/g' -e 's/{{MODULE}}/MODULE/g' \
  $SKILL_DIR/../pypi-claim/contents/pyproject.toml \
  > ~/.name-claim/$ARGUMENTS/pyproject.toml
sed 's/{{NAME}}/NAME/g' \
  $SKILL_DIR/../pypi-claim/contents/placeholder.py \
  > ~/.name-claim/$ARGUMENTS/MODULE.py
sed -e 's/{{NAME}}/NAME/g' -e '/^rm -rf /d' \
  $SKILL_DIR/../pypi-claim/contents/publish.sh \
  > ~/.name-claim/$ARGUMENTS/publish-pypi.sh
chmod +x ~/.name-claim/$ARGUMENTS/publish-pypi.sh

crate (if available):

sed -e 's/{{NAME}}/NAME/g' -e 's/{{TITLE}}/TITLE/g' \
  $SKILL_DIR/../crate-claim/contents/Cargo.toml \
  > ~/.name-claim/$ARGUMENTS/Cargo.toml
sed 's/{{NAME}}/NAME/g' \
  $SKILL_DIR/../crate-claim/contents/lib.rs \
  > ~/.name-claim/$ARGUMENTS/src/lib.rs
sed -e 's/{{NAME}}/NAME/g' -e '/^rm -rf /d' \
  $SKILL_DIR/../crate-claim/contents/publish.sh \
  > ~/.name-claim/$ARGUMENTS/publish-crate.sh
chmod +x ~/.name-claim/$ARGUMENTS/publish-crate.sh

Run the scaffolds for all available registries in the same Bash message (parallel). They write distinct paths so there are no race conditions.

5. Print the copy-paste block and stop

For every available registry, print one line in a single fenced bash block. The user copies the whole block and pastes it into their terminal as one chunk. && chaining would make each script depend on the previous one succeeding; if you want them independent the user can split. Default to newline-separated commands (independent execution):

Scaffolded: ~/.name-claim/$ARGUMENTS/

```bash
cd ~/.name-claim/$ARGUMENTS && ./publish-npm.sh
cd ~/.name-claim/$ARGUMENTS && ./publish-pypi.sh
cd ~/.name-claim/$ARGUMENTS && ./publish-crate.sh
```

Include only the lines for registries that were available. Skip the lines for registries that were taken or unknown.

Do NOT call Bash to run those commands. Do NOT print anything more about status. The user runs them and observes the results in their own terminal.

Prerequisites for publishing

The user handles auth in their own shell:

• npm: npm login (or NPM_TOKEN in env).
• pypi: API token in ~/.pypirc or TWINE_PASSWORD env var.
• crate: cargo login <token> from https://crates.io/settings/tokens.

If a publish fails, the user retries by hand from the same scaffold dir. The sister skills (/npm-claim, /pypi-claim, /crate-claim) remain available for one-off retries that start fresh.