The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/wispr.py utility uses subprocess.run to execute the system command pgrep to check if the Wispr Flow application is running. The command uses hardcoded arguments and does not incorporate user-provided input, preventing command injection.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes transcription data (speech-to-text results) which are untrusted external inputs. This creates a potential surface for indirect prompt injection if the transcripts contain instructions intended to influence the agent.
Ingestion points: Text data is read from the History table in scripts/wispr.py during history retrieval and counting operations.
Boundary markers: The skill instructions mandate that the agent must report its findings and obtain user approval before executing any modifications to the database.
Capability inventory: The skill is capable of reading from and writing to a local SQLite database and writing local JSON backup files.
Sanitization: All database interactions in the provided Python script use parameterized SQL queries, ensuring that transcribed text cannot be used to manipulate database command structures.

wispr-flow