The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external Excel files (.xlsx) via the tc-xlsx.py script. These files provide the test steps that the agent executes in a browser environment.
Ingestion points: The tc-xlsx.py extract command reads content from user-provided Excel files into tcs.json.
Boundary markers: The skill does not explicitly define prompt delimiters for the extracted test steps during browser execution.
Capability inventory: The agent has capabilities to control the browser (find, computer, read_page), write local result files, and call the tao_report_test MCP tool.
Sanitization: There is no explicit sanitization logic described for the test steps extracted from the spreadsheet.
[DYNAMIC_EXECUTION]: The skill generates execution plans and graphs at runtime.
Evidence: W02 involves generating a PLAN.md file and building an execution graph using orca-graph.py build based on the extracted test cases.
Risk: While structured, this represents dynamic assembly of instructions that guide the agent's behavior.
[COMMAND_EXECUTION]: The skill relies on executing local Python scripts to perform its primary functions.
Evidence: Execution of python3 harness/scripts/tc-xlsx.py and python3 harness/scripts/orca-graph.py to handle extraction, classification, and test orchestration.

tc-run