The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to maintain persistent context by reading from and writing to files in the .agents/ directory, which creates a potential surface for indirect prompt injection.
Ingestion points: The agent is instructed to read manager context from .agents/em-context.md and individual report profiles from .agents/reports/[name].md at the start of interactions.
Boundary markers: There are no explicit instructions or delimiters defined to ensure the agent ignores or sanitizes any embedded instructions within these context files.
Capability inventory: The skill allows the agent to automatically update these files based on conversation content, providing a file-write capability.
Sanitization: The skill lacks instructions for validating or sanitizing user-provided information before it is saved to the persistent context files.

developer-productivity