The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines roles that are instructed to execute local Python scripts. Specifically, the 'Product Manager' and 'User Researcher' roles in references/agent-registry.md utilize commands such as python scripts/rice_prioritizer.py features.csv --capacity 15 and python scripts/customer_interview_analyzer.py interview.txt to process data files.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to ingest and process untrusted external data (CSV/TXT files and user-provided text) through its collaborative agent workflows.
Ingestion points: references/agent-registry.md (reading features.csv and interview.txt via scripts) and SKILL.md (interpolating user-provided task descriptions).
Boundary markers: Absent. The instructions do not provide delimiters or specific guidelines for the agent to ignore potentially malicious instructions embedded within the processed data.
Capability inventory: The skill utilizes command execution for script processing and performs multi-step reasoning based on external inputs.
Sanitization: No evidence of sanitization, validation, or escaping of the external content was found in the documentation.

agent-team