The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided information (company details, problem statements, traction metrics) to populate a JSON file (pitch_data.json). This file is then consumed by an internal Python script to generate a PowerPoint deck. An attacker could potentially embed instructions within these fields to influence the agent's behavior during the information-gathering or file-generation phase.
Ingestion points: User input during the 'Gather Information' phase described in SKILL.md is written to pitch_data.json.
Boundary markers: The instructions do not define explicit delimiters or 'ignore embedded instructions' warnings when handling user data.
Capability inventory: The skill executes a local Python script scripts/create_pitch_deck.py and uses shell commands like grep on reference files.
Sanitization: There is no evidence of sanitization, escaping, or validation of the user-provided text before it is stored in the data file or processed by the generation script.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command: python3 scripts/create_pitch_deck.py pitch_data.json output_filename.pptx. The output_filename.pptx parameter is intended to be provided by the user, which presents a potential command injection surface if the agent does not properly validate the filename input before execution.
[EXTERNAL_DOWNLOADS]: The SKILL.md file recommends installing the python-pptx library via pip3 install python-pptx during troubleshooting. This is a standard dependency for the skill's functionality.

pitch-deck