j5ik2o-takt-skill-updater

Warn

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill clones the repository https://github.com/nrslib/takt.git using git to perform updates. This repository is not from a known trusted organization or vendor.
  • [COMMAND_EXECUTION]: The skill executes multiple shell commands including git clone, git checkout, rsync --delete, and runs local scripts such as scripts/sync-takt-references.sh and validate-order-md.sh that were not provided in the skill package for verification.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests instructional data from an external repository and then directs the agent to process this data to update its own core logic and instructions. This could be exploited if the external repository is compromised. Ingestion point is the git repository in SKILL.md, boundary markers are absent, and the agent has significant capabilities including file system modification.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 15, 2026, 03:33 PM
Security Audit — agent-trust-hub — j5ik2o-takt-skill-updater