The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local script aura-frog/scripts/preflight/run-all.sh using Bash to orchestrate a suite of security linters, including checks for YAML frontmatter, tool input shape, and ANSI-escape volume.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content such as file contents, shell commands, and stdin data for validation.
Ingestion points: Target resolution includes file paths, Bash command strings, and stdin content as specified in the Behavior section.
Boundary markers: There are no explicit markers or delimiters defined in the instructions to separate external input data from the agent's internal instructions.
Capability inventory: The skill is configured to execute Bash commands to perform its validations.
Sanitization: The skill implements its own security checks (e.g., check-secret-patterns.sh, check-path-safety.sh, and check-command-allowlist.sh) which act as a sanitization layer for the data it processes, mitigating the risk of executing or passing along malicious content.

preflight-validator