The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill repository includes a file named ppt_audit_result_v2.json which contains simulated security analysis results, such as 'verdict: safe' and 'security: PASS'. This is a self-referential content pattern intended to influence the judgment of automated security tools or human auditors.
Ingestion points: ppt_audit_result_v2.json contains static content meant to be read during analysis.
Boundary markers: None.
Capability inventory: The skill has file-writing capabilities through the python-pptx library and network fetching capabilities via urllib.
Sanitization: Not applicable to the static metadata file itself.
[EXTERNAL_DOWNLOADS]: The converter module contains functionality to fetch images from remote URLs for inclusion in generated slides.
Evidence: In framework/converter/components.py, the _fetch_image function uses urllib.request.urlopen to download content from a user-provided src URL.
[COMMAND_EXECUTION]: The skill provides several entry-point Python scripts (build_html.py, convert_to_pptx.py, validate_data.py) that perform file system operations and process user-defined JavaScript project files using regular expressions and JSON parsing.
Evidence: build_html.py uses regex and file reads to inline CSS and JS; convert_to_pptx.py invokes the slide rendering engine to write .pptx files to the local output/ directory.

ppt