Skills
Navigate
skills/ovachiever/droid-tings/speculative-decoding/Gen Agent Trust Hub

speculative-decoding

Warn

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires downloading code from GitHub repositories that are not on the trusted vendor list.
  • Evidence: git clone https://github.com/FasterDecoding/Medusa in SKILL.md.
  • Evidence: git clone https://github.com/hao-ai-lab/LookaheadDecoding in SKILL.md.
  • [REMOTE_CODE_EXECUTION]: The skill executes code from external repositories during the installation phase.
  • Evidence: pip install -e . is used to install the cloned Medusa and LookaheadDecoding repositories.
  • [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted user input in prompt generation without sanitization.
  • Ingestion points: prompt variable in SKILL.md is passed to model inference functions.
  • Boundary markers: None present.
  • Capability inventory: Inference capabilities provided; no dangerous tools are linked to model outputs.
  • Sanitization: None.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 8, 2026, 09:45 PM