The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions in SKILL.md and artifact-contract.md direct the agent to execute specific Python scripts bundled with the skill, such as workbook_inventory.py and profile_reconcile.py, to analyze local spreadsheet files.
[DYNAMIC_EXECUTION]: The skill uses local Python scripts to process user-supplied data at runtime. These scripts are developed with defensive programming practices, including resource limits (MAX_ROWS, MAX_INPUT_BYTES) and XML safety checks to prevent attacks like XML External Entity (XXE) and resource exhaustion when inspecting OOXML workbooks.
[EXTERNAL_DOWNLOADS]: The references/tooling.md file describes an optional integration using @negokaz/excel-mcp-server via npx. This highlights a potential download of external code from the npm registry if the operator manually chooses to enable this feature.
[PROMPT_INJECTION]: The file eval-spec.yaml contains an adversarial prompt ("Ignore all previous instructions...") within a test_cases definition. This is used exclusively for evaluating the agent's resilience during testing and does not influence the agent's behavior during normal operation.

excel-analyst-pro