[25 MRG] Sign pack: `vsl_school` (vsl) — photo/video + landmarks

[TUPM96]

Bounty: 25 MRG — gloss / gesture pack

Add isolated sign vsl_school to Loru so sign→text / sign→voice can learn it.

Field	Value
Gloss id	vsl_school
Category	vsl
Description	VSL school
Handshape (hint)	document in PR
Motion (hint)	document in PR

Hints above are starting points (often ASL-like). If you contribute VSL or another language, document the exact form in the PR (and set language in the sample JSON). Regional variants are welcome when consented and labeled.

What to deliver (PR)

1. Sample sequence JSON — data/samples/vsl_school.json (or data/samples/<lang>/vsl_school.json)

   • gloss: vsl_school
   • language: e.g. demo-asl, vsl, asl
   • fps, source: contributor-capture + your handle
   • frames: landmark frames (hands; pose optional)
     • Prefer MediaPipe-style lists of [x,y,z] per landmark
     • Or synthetic but distinct prototype if you cannot run vision yet — mark source honestly
   • Optional: notes describing handshape/path/NMM

2. Register gloss — extend DEFAULT_GLOSS in src/loru/models/vocab.py or add to a new vocab file if the project already supports external vocab (follow current code).

3. Sentence template — add a natural-language string for this gloss in gloss_to_sentence (or multi-gloss pack).

4. Evidence (required) — in the PR body:

   • ≥ 2 still photos of the sign (start pose + end / peak pose), or
   • 1 short video clip (≤5s) of the isolated sign
   • Must be original / consented. No scraped private faces without permission.
   • Blur faces if the signer prefers privacy (hands+torso enough).

5. Verify

   pip install -e ".[dev]"
   # after adding sample:
   loru data list
   loru infer text --sequence data/samples/vsl_school.json
   pytest -q
   ruff check src tests

Acceptance

• Sample JSON merged with correct gloss
• Vocab + text template updated
• Photo/video evidence linked in PR
• Consent / ethics note in PR (who signed; permission)
• Tests + ruff pass
• Language labeled (asl / vsl / other)

Ethics

• Only consented signers (yourself OK).
• Do not claim clinical accuracy for assistive use without evaluation.
• Prefer open documentation of which sign language variety you captured.

Claim (MergeOS MRG)

1. Star https://github.com/mergeos-bounties/Loru and https://github.com/mergeos-bounties/mergeos
2. Comment on this issue: I claim this bounty
3. Comment on MergeOS Claim Token #1 with this issue link
4. Open a PR to Loru with Fixes #<this-issue>

Policy: docs/BOUNTY.md

Payout

Maintainer reviews PR → merge → MRG credit (typically 25 MRG per gloss pack on the 25/50/100/200 scale).

[yh-liao-07]

I claim this bounty

[laurentketterle-hub]

I claim this bounty. PR ready for review: #593

[laurentketterle-hub]

/attempt 183

[laurentketterle-hub]

/claim