tone-manner-maker
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted writing samples from various sources, which is a known vector for indirect prompt injection attacks. \n
- Ingestion points: Untrusted data is ingested through text pastes, uploaded files, and the content of external URLs (as specified in
SKILL.md).\n - Boundary markers: The skill employs clear boundary instructions in the
<untrusted_samples>section ofSKILL.mdand the 'Sanitize rules' inrules/sample-quality.md, explicitly telling the agent to treat samples as data only.\n - Capability inventory: The skill is restricted to text analysis and Markdown generation, with no access to system commands or external network writing tools.\n
- Sanitization: The agent is instructed to strip non-prose content (boilerplate, comments) from URLs and ignore imperative text within any sample.\n- [PROMPT_INJECTION]: Static analysis detected instructions related to ignoring previous commands in
SKILL.mdandrules/sample-quality.md. These are confirmed to be defensive safety filters designed to prevent the agent from executing malicious commands embedded within user-provided samples, rather than an attempt to bypass system safety guidelines.\n- [COMMAND_EXECUTION]: The skill does not contain any executable scripts or commands. It relies purely on Markdown-based instructions and local file references. References tonodeorbunin the validation section are for development-time consistency checks and not part of the runtime execution path.
Audit Metadata