The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/download_slides_pages_voices.py is vulnerable to path traversal. The --job-id command-line argument is used directly to construct the output file path (f"{job_id}.zip") without any sanitization or validation. This allows a potentially malicious job ID containing directory traversal sequences (e.g., ../../etc/passwd) to result in files being written to arbitrary locations on the host system, depending on the environment's permissions.\n- [INDIRECT_PROMPT_INJECTION]: The skill's document summarization feature introduces an attack surface for indirect prompt injection.\n
Ingestion points: The skill ingests content from external PDF, DOCX, and TXT files as described in the summarization workflow in SKILL.md.\n
Boundary markers: None. The skill does not define specific delimiters or instructional guardrails to prevent the AI agent from following malicious instructions that might be embedded within the documents being processed.\n
Capability inventory: The skill possesses the ability to perform network operations (POST requests to the 2slides API) and write files to the local filesystem.\n
Sanitization: None. Content extracted from external documents is structured into slide content without being sanitized or escaped, which could influence downstream agent behavior.\n- [EXTERNAL_DOWNLOADS]: The skill interacts with the 2slides.com API to generate presentations and narrations. It includes functionality to download generated assets as ZIP archives from URLs provided by the service's API, which is consistent with its stated purpose.

2slides