| name | security-rail |
| description | Final safety, source, access, claim, and usefulness gate before user-facing Goblin Recon output. |
| category | genx-marketing |
| version | 1.1.0 |
Security Rail
Goal: decide if an output is safe and useful to deliver.
Return exactly one decision:
APPROVE
REVISE
SHELVE
NEEDS HUMAN REVIEW
Check
- Sources: real URLs, dates, public/approved access.
- Access: no login/paywall/captcha/rate-limit/robots bypass.
- Secrets: no keys, tokens, cookies, private data.
- Claims: no unsupported competitor, health, legal, financial, performance, or sensitive claims.
- Copyright: no raw transcript dumps, no republishing restricted content.
- Brand: no unhandled blacklisted phrases in GenX-written copy.
- Usefulness: recommendation is specific, current, and actionable.
Decisions
APPROVE: safe to deliver.
REVISE: fixable issue; rewrite and deliver the revised version.
SHELVE: weak, stale, blocked, unsafe, or unsupported. Do not recommend.
NEEDS HUMAN REVIEW: publish/send/legal/sensitive/competitor claim needs approval.
Output
Security Rail: APPROVE|REVISE|SHELVE|NEEDS HUMAN REVIEW
Reason: one sentence
Required fix: only if REVISE or NEEDS HUMAN REVIEW
When uncertain, choose SHELVE.