skills/tmeister/skills/vault-session/Gen Agent Trust Hub

vault-session

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local git, ls, grep, and mkdir commands to gather project context and manage notes within a hardcoded local vault directory (/Users/tmeister/Vault). The commands are used legitimately to automate note creation and update workflows.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting untrusted data from the local git repository—such as branch names, commit messages, and file diffs—and using it to reconstruct session narratives.
  • Ingestion points: Git repository metadata via git branch, git log, and git diff in SKILL.md (Workflow Step 1).
  • Boundary markers: None present; the raw git output is processed and distilled directly into the note template.
  • Capability inventory: File system write access via mkdir and file creation/update logic in SKILL.md (Workflow Step 6).
  • Sanitization: No sanitization or validation is performed on the data retrieved from git before it is used to generate the session summary.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 08:02 PM
Security Audit — agent-trust-hub — vault-session