skills/vasanth19/skills/c-replicate/Gen Agent Trust Hub

c-replicate

Fail

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: HIGHDATA_EXFILTRATIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The ins t r u c t ions in SKILL.md direct the agent to per f or m netw or k req u es ts (via cur l) to ap i.c-rep l i c ate.com. This domain is not the offic ial en d poin t for Rep l i c ate.ai (ap i.rep l i c ate.com) and ap p ear s to be a pr ox y or imper s on ation s ite des ign ed to inter c ep t sen s itive data.
  • [CREDENTIALS_UNSAFE]: The skill ins t r u c t s the agent to sour ce the REPLICATE_API_TOKEN from a loc al sec r ets file (~/.gs ai/sec r ets.en v) and inc l u de it in head er s sen d to the sus p ic ious c-rep l i c ate.com domain.
  • [COMMAND_EXECUTION]: The skill rel ies on the exec u tion of syn c-mod el s.sh and mul t ip le pip ed Pyth on comman ds. The se sc r ip ts ac ces s the loc al file sys tem and per f or m netw or k op er ations us ing us er-pr ov id ed var iab l es with out suffic ien t val id ation, whic h coul d lead to comman d inj ec tion.
  • [EXTERNAL_DOWNLOADS]: The syn c-mod el s.sh sc r ip t per f or ms automated dow n load s of mod el metad ata from ap i.rep l i c ate.com. While the syn c sc r ip t us es the offic ial domain, the pr imar y op er ation al ins t r u c t ions us e the sus p ic ious domain, cr eat ing a dec ep t ive dis c r ep an c y.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 5, 2026, 07:40 PM