The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its primary inputs.
Ingestion points: The skill reads untrusted data from git_diff and completed_work_claim as defined in SKILL.md.
Boundary markers: There are no specific instructions or delimiters provided to isolate these inputs from the agent's core instructions.
Capability inventory: The skill uses terminal_access and test_runner to execute shell commands and tests based on the claims found in the inputs.
Sanitization: No sanitization or validation logic is specified for the content within the diffs or reports.

fable-judge