kb-check-drift

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes git log and git diff to analyze repository history and identify changes since the last recorded refresh. It also uses grep to cross-reference these changes with existing knowledge base indexes and summaries.
  • [COMMAND_EXECUTION]: Uses the mv command to perform atomic writes for its internal report artifact (.kb/.drift-report.json), ensuring data integrity during the update process.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes untrusted data such as git commit messages, file paths, and file contents to generate refresh recommendations. While it lacks explicit boundary markers, it only generates human-readable recommendations and does not auto-execute commands based on the data.
  • Ingestion points: .kb/.kb-state.json, git history metadata (logs, diffs), and repository file contents (summaries, indexes, plans).
  • Boundary markers: None identified; data is treated as plain text for analysis.
  • Capability inventory: Shell command execution (git, grep, mv, ls) and local file system access.
  • Sanitization: None identified; analysis relies on direct output from system tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 07:19 PM
Security Audit — agent-trust-hub — kb-check-drift