The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user requests to modify source code and project documentation, creating a surface for indirect prompt injection.
Ingestion points: The skill reacts to user prompts such as feature requests or bug fix instructions (e.g., 'Añade exportación a CSV').
Boundary markers: The workflow requires the creation of a 'Feature Document' in 02-DOCS/wiki/ftd/ before any changes are made. This document serves as a structured boundary defining intent and scope.
Capability inventory: The skill is designed to manage documentation and request branch creation or unlocking via environment-specific tools like rsc.
Sanitization: The instructions do not specify automated sanitization but rely on the agent following a strict evidence-based checklist protocol.
[COMMAND_EXECUTION]: The skill references environment-specific commands such as rsc main unlock and rsc/knowledge. These appear to be internal tools belonging to the project harness rather than arbitrary shell execution or malicious payloads.
[DATA_EXPOSURE]: The skill reads from a user profile document at 02-DOCS/wiki/harness/user-profile.md to determine the appropriate technical level for responses. This is a local configuration access consistent with the skill's stated purpose of providing oriented guidance.

ftd