The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external datasets (such as QTO quantities and pricing rules) provided in CSV, Excel, or tabular formats, which creates a surface for potential instruction injection if the data contains text intended to override agent behavior.\n
Ingestion points: instructions.md (Parsing logic for CSV, Excel, and tables) and SKILL.md (methods load_pricing_rules_from_df and load_qto_from_df).\n
Boundary markers: The instructions lack explicit delimiters or warnings to treat the ingested data solely as data and ignore any embedded natural language instructions.\n
Capability inventory: The skill utilizes filesystem permissions in claw.json and performs file-writing operations via the pandas library in SKILL.md.\n
Sanitization: No input sanitization or validation logic is present to filter or escape instructions embedded within the quantity or rule data.

auto-estimate-generator