The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes previously created markdown files in 'Update' mode, which introduces a surface for indirect prompt injection if those files are modified to include malicious instructions.
Ingestion points: Existing OODA documents located in project directories like decisions/, strategy/, and product/.
Boundary markers: There are no explicit delimiters or instructions to the agent to ignore potentially malicious content within the files it reads.
Capability inventory: The skill can write to the filesystem and call other skills such as /decision-record or /strategic-bet.
Sanitization: No sanitization or content validation is performed on the files retrieved from the local environment.

ooda-loop