asciinema-recorder

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFEPROMPT_INJECTIONPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains 'Self-Evolving' instructions that mandate the agent to modify its own source code (SKILL.md) upon encountering issues or drift. This creates a mechanism for persistent instruction modification that could be subverted.
  • [PRIVILEGE_ESCALATION]: The installation phase includes the use of sudo apt install for Linux systems, which requests elevated privileges to install the asciinema package.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to perform environment discovery (e.g., basename "$PWD", date) and setup tasks (e.g., mkdir -p). It also provides the user with a generated shell command for terminal recording.
  • [INDIRECT_PROMPT_INJECTION]: The skill incorporates external data from the file system and user prompts into its operational logic.
  • Ingestion points: The current working directory path ($PWD) and user-supplied strings for recording titles and paths via AskUserQuestion.
  • Boundary markers: There are no explicit delimiters or safety instructions used when interpolating these variables into shell script blocks.
  • Capability inventory: The skill utilizes the Bash tool for directory creation and system information retrieval, and it can execute package managers.
  • Sanitization: The basename command is used to extract the workspace name, providing some structural validation, but other inputs (like custom titles) are interpolated directly into command templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 06:30 AM