The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources and user inputs, which creates a potential surface for indirect prompt injection.
Ingestion points: The skill reads data from fetched web pages, uploaded CRM exports, and user-pasted notes as defined in Step 1 and Step 2 of the workflow in SKILL.md.
Boundary markers: SKILL.md includes explicit instructions for the agent to treat fetched pages and user-pasted content as data only, with a directive to ignore any instructions found within that data.
Capability inventory: The skill executes local Python scripts (scripts/icp_gap.py and scripts/fit_count.py) via subprocess calls and has access to network-enabled tools (web_search, web_fetch) as described in the operating-principles.md reference.
Sanitization: The skill relies on natural language instructions to separate data from commands but does not implement programmatic sanitization or strict schema validation for the raw text inputs before processing.

icp-review