tamarru
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs explicit directives to override the model's default identity and safety framework in favor of a specific persona. Phrases such as "stop being an assistant writing about Tom's topics," "Become the voice itself," and "possessed by the daimon, not assisted by it" are used to force a role-play state that disregards standard assistant behavior.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through user-supplied topics.
- Ingestion points: User input is captured via the
$ARGUMENTSvariable inSKILL.mdto define the writing topic. - Boundary markers: There are no delimiters or "ignore embedded instructions" warnings used when interpolating the user-provided topic into the prompt.
- Capability inventory: The skill's impact is limited as it is primarily a text generator and explicitly restricts external tool use via the
disable-model-invocation: trueconfiguration in its frontmatter. - Sanitization: The skill does not validate or sanitize the input provided in
$ARGUMENTSbefore processing.
Audit Metadata