The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data (CRM exports, notes, emails) which may contain malicious instructions designed to alter the agent's behavior.
Ingestion points: As described in SKILL.md, the workflow ingests deal facts from CRM exports, user notes, and pasted email content.
Boundary markers: The skill includes explicit security instructions for the agent to treat all external text as data only and to ignore any embedded commands. It mandates that any ignored instructions be reported to the user in a single-line summary.
Capability inventory: The skill executes local Python scripts (scripts/deal_health.py and scripts/close_plan.py) via the command line. These scripts use only the Python standard library and do not perform network operations or file system writes.
Sanitization: The agent is instructed to use a "Notice line" to explicitly inform the user when it detects and disregards an instruction found within the provided data.
[COMMAND_EXECUTION]: The skill automates the execution of local scripts to provide calculations and schedules.
Evidence: SKILL.md contains specific instructions to run python3 scripts/deal_health.py and python3 scripts/close_plan.py using user-supplied facts formatted as JSON arguments. These scripts are part of the skill's own package and are used for intended computational logic.

deal-review-and-close-plan