pitch-deck-builder-ai-labs
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes user-supplied business information to generate presentation content.
- Ingestion points: Business details (company name, problem, solution, traction, financials) gathered from the user during the initial workflow steps defined in "SKILL.md".
- Boundary markers: None explicitly used in the instructions to separate user-provided data from agent instructions or to warn against embedded malicious prompts.
- Capability inventory: The skill allows the agent to execute a local Python script ("scripts/create_pitch_deck.py") and read/write local files ("pitch_data.json" and ".pptx" output files).
- Sanitization: Data is handled as structured JSON, which provides basic structural separation, but the content is eventually rendered into PowerPoint slides without specific filtering for malicious instruction strings.
- [COMMAND_EXECUTION]: The skill workflow relies on executing shell commands to process data and generate the final document.
- Evidence: "SKILL.md" directs the agent to execute "python3 scripts/create_pitch_deck.py" with user-defined input and output filenames. Additionally, it uses "grep" to retrieve best practices from the "references/pitch_deck_best_practices.md" file.
Audit Metadata