The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external feature specifications (PRDs, user stories) and JSON-formatted scenario lists, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through the analysis of feature specifications in the workflow and the processing of feature_spec.json by the auditing script.
Boundary markers: The skill does not define specific delimiters or instructions to prevent the agent from obeying potentially malicious instructions embedded in the user-provided specifications.
Capability inventory: The skill includes a local Python script (test_scenario_generator.py) for auditing coverage. This script has no network capabilities and only interacts with the local file system to read input and optionally write reports.
Sanitization: The skill does not implement sanitization or validation of the text content within specifications to filter out potential injection attempts.
[SAFE]: While the skill's reference materials and instructions frequently mention security vulnerabilities such as SQL Injection (SQLi) and Cross-Site Scripting (XSS), these are used purely as educational examples for defining test coverage and do not represent malicious instructions or a prompt injection attack.

test-scenarios