talk-script-clarity
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text data in the form of presentation scripts or speaker notes. If these inputs contain hidden instructions, they could potentially influence the agent's behavior.
- Ingestion points: The agent reads user-provided scripts in markdown or HTML formats, and these are also passed to the
clarity_metrics.pyscript for analysis. - Boundary markers: The skill recommends using specific slide headers (e.g.,
[번호 · 제목 · 초]) to structure the input, which provides some structural context but does not fully isolate the content from instruction processing. - Capability inventory: The skill possesses the capability to execute a local Python script and read from the file system to perform metric calculations.
- Sanitization: The bundled metrics script uses regular expressions to extract data but does not perform sanitization to strip potential prompt injection sequences from the text before it is processed by the LLM.
- [COMMAND_EXECUTION]: The skill's operational workflow includes instructions for the agent to execute a local Python utility (
scripts/clarity_metrics.py). The script is used for text parsing and metric calculation (e.g., sentence length, jargon counting) and does not perform network operations or access sensitive system files.
Audit Metadata