The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external quantity snapshots and project-related files, which constitutes a potential attack surface for indirect prompt injection.
Ingestion points: The skill reads external source snapshots and several local project files including PROJECT.md, STUDIO.md, TASKS.csv, ffe/items-and-schedules, and product-library.csv as defined in host-contract.json.
Boundary markers: There are no explicit instructions or delimiters mentioned in SKILL.md to help the agent distinguish between data and instructions within the processed files.
Capability inventory: The skill is configured with Read, Write, and Bash tools. The availability of a shell (Bash) alongside the processing of untrusted data inputs increases the potential impact of an injection.
Sanitization: Instructions focus on data reconciliation logic (unit conversion, duplicate key resolution) but do not include measures to sanitize or escape the content of input files to prevent instruction override or command injection.

schedule-quantity-reconcile