The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its handling of external project data and user input.
Ingestion points: Data is pulled from business cases (/business-case), strategic bet records in context/bets/, and user-supplied signals during the evaluation process as described in the Mode Behaviors section of SKILL.md.
Boundary markers: The instructions lack directives to use delimiters or specific boundary markers when interpolating content from the continuation_threshold field of business cases into the evaluation output.
Capability inventory: The skill performs file read and write operations across context/ and reviews/ directories to update bet audit trails and save evaluation artifacts.
Sanitization: There is no evidence of filtering or validation of the content from the business cases, as the skill is instructed to restate threshold fields verbatim in the output structure.

continuation-threshold