presenting-conference-talks
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from research papers to generate presentation content, which could lead the agent to follow malicious instructions hidden within the document text.
- Ingestion points: Research paper files (PDF or LaTeX source) as defined in the Workflow section of
SKILL.md. - Boundary markers: Absent. No delimiters or instructions are provided to the agent to treat paper content strictly as data and ignore any embedded commands.
- Capability inventory: The skill instructs the agent to execute shell commands (
latexmk,python3) and write files (slides.tex,talk.pptx). - Sanitization: Absent. There is no evidence of validation or filtering of the content extracted from the papers before it is used in slide generation.
- [DYNAMIC_EXECUTION]: The workflow requires the agent to generate a Python script from a template and subsequently execute it to produce the final PPTX output.
- Evidence: Step 4 of the workflow in
SKILL.mdexplicitly calls for runningpython3 generate_slides.py. - Context: The script generation is based on a transparent template in
references/slide-templates.mdand uses the well-knownpython-pptxlibrary for standard document creation tasks.
Audit Metadata