skills/cloudposse/atmos/atmos-cast/Gen Agent Trust Hub

atmos-cast

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill enables recording terminal sessions, including stdin keystrokes when the input: true configuration is enabled. This presents a risk of sensitive data exposure where passwords, API keys, or private local paths could be captured and persisted in .cast files or rendered artifacts (GIF, MP4, etc.).
  • [COMMAND_EXECUTION]: The skill facilitates the execution of shell commands via the atmos CLI. It describes methods to trigger recording and execution through CLI flags, environment variables like ATMOS_CAST, and workflow step definitions.
  • [DYNAMIC_EXECUTION]: The skill supports executing Starlark scripts for recording validation (interpreter: starlark). This allows for runtime data processing and logic execution based on the contents of the recorded terminal sessions.
  • [INDIRECT_PROMPT_INJECTION]: By processing and rendering terminal output that can be influenced by external factors (such as the output of third-party tools or file names on disk), the skill exposes a surface for indirect prompt injection. This risk is present if an agent is subsequently directed to interpret or follow instructions found within the generated recordings.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 07:52 AM