The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data (database schemas and JSON API responses) which creates a surface for indirect prompt injection attacks.
Ingestion points: Reads developer files including schema.prisma, models.py, db/schema.rb, migration files, and the app/Models directory, as well as JSON response samples from APIs.
Boundary markers: No explicit boundary markers or delimiters are instructed to be used when the agent processes these files.
Capability inventory: The skill can execute shell commands via the endpoint CLI, write files to the local filesystem (propuesta.json), and perform network requests (endpoint submit).
Sanitization: The skill implements a validation step via endpoint validate which is designed to block credentials and ensure profile compliance before submission.
[COMMAND_EXECUTION]: The skill relies heavily on a local CLI tool named endpoint to perform all core logic, including system detection, mapping, and data submission.
[DATA_EXFILTRATION]: The skill includes a submission step (endpoint submit) that sends a generated proposal to an external API. The instructions explicitly mandate that only authentication header names be included, never the keys themselves, and requires user confirmation for public contact information.

expose-endpoint