tk-artifacts
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a standard for ingesting user-provided data via JSON files, establishing a vulnerability surface for processing external content.
- Ingestion points: User-edited JSON files located in temporary (.tigerkit/tmp/) and secret-input (.tigerkit/secret-input/) directories.
- Boundary markers: Instructions mandate strict JSON schema validation and prohibit the use of comments, JSON-LD, or instructions within the data files.
- Capability inventory: The skill references shell commands (git, rm) and a local script (scripts/durable_artifacts.py) for managing artifacts.
- Sanitization: Instructions require validating keys and types before consumption and strictly forbid exposing raw input or parser errors to the agent's output.
- [COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to verify repository state and perform safe file deletions.
- Evidence: Mentions the use of git ls-files, git check-ignore, and rm commands, with strong emphasis on using literal path arguments to prevent command injection or accidental deletions.
Audit Metadata