The Agent Skills Directory

[SAFE]: No security vulnerabilities were detected. The skill is entirely instructional, focusing on response style and the elimination of sycophancy.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by processing user-provided queries and drafted responses to produce formatted output. However, it lacks any exploitable capabilities to execute instructions found in that data.
Ingestion points: SKILL.md defines raw_response and user_query as primary inputs.
Boundary markers: Absent; the skill relies on the model to interpret and transform the inputs based on the outcome-first policy.
Capability inventory: None. The skill.package.json specifies scriptPolicy: "data-only" and there are no scripts or tool invocations in the procedure.
Sanitization: Not present, though the risk is negligible due to the lack of dangerous tools.

fable-outcome-first