capture-a-prompt
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill workflow facilitates the permanent storage of untrusted user input into a prompt library. These instructions are then available for reuse by the agent as slash commands, which could lead to persistent malicious behavior if the ingested text contains prompt injections.
- Ingestion points: User-pasted text captured in the primary workflow (SKILL.md).
- Boundary markers: Absent; the skill lacks delimiters or instructions to ignore embedded commands, and explicitly mandates that the agent 'Keep their wording'.
- Capability inventory: File system write access to the
commands/directory andINDEX.md, and execution of shell scripts. - Sanitization: None; the agent is instructed to avoid wrapping input in extra headings or editing content beyond basic metadata.
- [COMMAND_EXECUTION]: The skill requires the execution of a local script
./scripts/install-local.shto register the new command into the environment, representing a standard but untrusted command execution pattern.
Audit Metadata