| name | ste100 |
| description | Local-only ASD-STE100 Simplified Technical English gate for Grok Build. No API keys, no network, no external services. Pure skill instructions: orchestrator + editor + adversarial rubric loop. Triggers on docs, procedures, runbooks, STE, simplified English, ASD-STE100, maintenance steps.
|
| metadata | {"short-description":"Local STE100 plugin — no API calls"} |
STE100 (Grok Build) — local plugin, zero API calls
This skill is instruction-only. Do not call the xAI API, HTTP endpoints,
or any network STE checker while enforcing these rules. Compliance is performed
in-context using rules.md + loop.md.
Non-negotiable
- No API calls
- No secrets / env keys required
- Edit only workspace documentation text
- Simulate orchestrator → editor → adversary locally
Roles
- Orchestrator — rounds, accept/stop, best candidate
- Editor (doer) — rewrite from original source at level 1→3
- Adversary — rubric + hard gates
Loop
baseline adversary(original)
while not pass and round <= 3:
candidate = editor(original, level=round)
report = adversary(candidate)
keep best
return best
Never chain editor→editor on a prior draft. Escalate by re-rewriting the source.
Rubric floors
Target score 92. Axis floor 80. Any error = FAIL.
Axes: length 20% · one instruction 18% · voice/verbs 18% · vocabulary 20% ·
form 14% · consistency 10%.
Editor levels
| L | Work |
|---|
| 1 | Vocabulary / full forms |
| 2 | Structure + imperative |
| 3 | Full STE rebuild |
Rules
See rules.md and loop.md.
Grok Build UI note
When building STE tooling UIs, prefer local deterministic demos. Do not wire
visitor-facing LLM API calls for this skill.