The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data via its market classification tool.
Ingestion points: The scripts/market_segment_classifier.py script reads a local text file supplied as a CLI argument (e.g., description.txt).
Boundary markers: The skill does not implement delimiters or explicit guardrails to prevent the agent from being influenced by malicious instructions embedded within the processed business descriptions.
Capability inventory: The script performs local file system reads and regex-based text matching; it contains no network or file-write capabilities.
Sanitization: No sanitization or filtering is performed on the input text before the agent is instructed to use the resulting classification for strategic guidance.
[COMMAND_EXECUTION]: The core workflow of the skill involves executing the provided Python script scripts/market_segment_classifier.py within the local environment.
[DATA_EXFILTRATION]: When the --json flag is used, the classification script prints the first 200 characters of the input file. This allows for localized data exposure; if a user or malicious prompt induces the agent to run the classifier on sensitive files (such as SSH keys or environment variables), the file's header will be revealed in the tool's output.

proptech-advisor