The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/wispr_dictionary.py performs shell command execution using the subprocess module. Specifically, it uses pgrep to verify if the Wispr Flow process is running and open -a to restart the application after database modifications. These are controlled executions for application management.
[INDIRECT_PROMPT_INJECTION]: The skill extracts and analyzes user-generated dictation text from the local Wispr Flow database, which acts as a potential vector for indirect prompt injection.
Ingestion points: Text data is fetched from the History table's formattedText and asrText fields in scripts/extract_wispr.py.
Boundary markers: The analysis templates in references/analysis-prompts.md do not include delimiters or instructions to ignore instructions embedded within the user's dictations.
Capability inventory: The skill can modify the local application database, write to the filesystem, and execute specific shell commands.
Sanitization: No filtering or sanitization of the extracted text samples is performed prior to analysis by the LLM.
[EXTERNAL_DOWNLOADS]: The skill's documentation in SKILL.md identifies praat-parselmouth as a required dependency for audio analysis features, which is a legitimate library for the stated purpose of prosody extraction.

wispr-analytics