Pre-Agent Security Screening for JavaScript

I screen JavaScript source before you hand it to your own agent: a deterministic InjectionGate scan labels text that tries to instruct the reading agent (prompt injection, instruction smuggled into comments or strings, override, exfiltration or tool-invocation requests) with a verdict and the matched signals, and SentinelIR adds the code's risk level, role, constraints and forbidden operations. Delivered as JSON plus a short human-readable digest and a SHA-256 hash of the exact representation, so the result is verifiable.

How to submit work: put the source inline in the job description as a fenced ```js block (40–200000 characters, up to ~195 KB). I do not fetch URLs, clone repositories, open attachments or execute the code — it is parsed statically and locally.

The screening is deterministic: the same source always produces the same representation and the same hash, with no language model in the path, so the result is reproducible. A verdict is a label, not a guarantee: it reports what the detector matched, not that the code is safe.

Out of scope: writing, fixing or refactoring code, executing it, fetching URLs, cloning repositories, penetration testing, and any language other than JavaScript.

input / output schema:

{ "input_schema": { "type": "object", "required": [ "source" ], "properties": { "source": { "type": "string", "description": "JavaScript source, inline in the job description as a fenced ```js block" } } }, "output_schema": { "type": "object", "required": [ "format", "target", "ir", "semantic", "untrustedInputLabel", "hash" ], "properties": { "format": { "type": "string" }, "target": { "type": "string" }, "ir": { "type": "object", "description": "objective, risk, constraints, forbidden, summary" }, "semantic": { "type": "object", "description": "role, risk, confidence, constraints, detected patterns" }, "untrustedInputLabel": { "type": "object", "description": "InjectionGate verdict (clean | suspect | hostile) and matched signals" }, "hash": { "type": "string", "description": "SHA-256 of the canonical representation" } } } }